tension-leg platform (TLP) rigid riser design considerations and presents results of TLP riser analysis for a pro- duction/injection riser in the North Sea. Analysis methods, design cri- teria, and design optimization are ad- dressed. The riser is designed for 300-m water depth in compliance with Norwegian regulations. Empha- sis is placed on application of the regulations and quantitative compar- ison of alternative methods for analysis of fatigue and extremes. 326 TLP Rigid Riser: A Case Study P.P. Rooney* and K.B. Engebretsen, Aker Engineering A/S, and D.J. Pettersen, Saga Petroleum A/S Introduction This paper summarizes TLP rigid riser de- sign considerations and presents results of the TLP riser analysis. TLP functions in- clude drilling, production/injection, and ex- port through rigid risers. TLP design is most sensitive to the many production/injection risers. Such parameters as riser spacing, top tension, stroke, and Pontoon clearance strongly influence global dimensions and deck spacing/layout. Parameters used in this case study are summarized in Table 1. Fig. 1 is a schemat- ic of the TLP riser system. The design is based on Norwegian Petroleum Directorate regulations 1 and guidelines. 2 Criteria for local design are summarized in Ref. 3. Analysis Methods Both frequency domain and time domain analysis methods 4 ,5 were used. Fatigue analysis is primarily based on frequency do- main and extremes on time domain. The riser is a nonlinear dynamic system, and the inaccuracy in using frequency domain (as- suming constant top tension and using stochastic linearization) is discussed below. The approach adopted for extreme responses is regular wave analysis. This method was used to overcome uncertainty in statistical extrapolation from an irregular time simulation and, for efficiency, to enable many design iterations. Dynamic analysis was carried out without introducing load factors. Load factors are applied to responses when carrying out code checks and when preparing design values for component specifications (e.g., stroke results). Boundary conditions include TLP motion, tensioner characteristics (see Stroke section) and template/subsea wellhead interface (see Components and Interfaces). The riser anal- ysis included TLP offsets in the range cor- responding to TLP extreme motions. 6 First-order response is included directly as a transfer function in the riser analysis, and all other TLP motion components are taken into account by defming a range of riser mean positions upon which the first-order 'Now with Conoco Norway Inc. Copyright 1992 Offshore Technology Conference response is superimposed. TLP setdown is included at each timestep in the time domain analysis. Sensitivity and Optimization As usual for deterministic design, an exten- sive sensitivity study must be carried out. The following were investigated: (1) wave periods; (2) current velocity; (3) riser loca- tion and geometric phase; (4) hydrodynamic parameters, variation with depth, Reynold's number,1, Keulegan-Carpenter's number, 1 and roughness; (5) annulus and tubing fluids; (6) top tension and tensioner stiffness characteristics; (7) TLP offset; (8) extent of marine growth; and (9) TLP hull influence on wave kinematics. Quantitative sensitivity results vary widely for the various responses along the riser, and the resulting priority is influenced by those responses that govern the design. The rigid risers tend to be in conflict with the overall TLP optimization on such param- eters as top tension and riser spacing. Op- timization, therefore, is an iterative process carried out on the basis of the sensitivity re- sults for both the risers and the TLP. Handling and operational weather limita- tions are most sensitive to riser spacing and wellbay layout. These require early detailed consideration because they are important to TLP efficiency in service. An alternative analytical approach to sen- sitivity analysis and optimization with probabilistic methods could be used where quantitative sensitivity results are generated directly in one analysis. This method has been applied successfully to TLP tethers. 8 Extreme Stresses Allowable tensile stresses in the riser wall were checked against values in Ref. 2. The Von Mises (combined) stresses were solved with a minimum wall thickness, taking into account corrosion/wear allowance and wall- thickness tolerances. Ref. 9 is used for stress calculations. Compressive stresses also were checked for load cases where axial stresses were low. Local buckling checks did not lead to addi- tional design requirements. Extreme results vary considerably with analysis method. An example is given for March 1992 JPT TABLE 1-SUMMARY OF PARAMETERS USED IN OTC RISER CASE STUDY Case study data Mean sea level, m Bottom of riser model. m Riser spacing, center to center, m Top of riser model, m Pressure at top of riser, MPa Top tension (including 1()()..kN surface wellhead and tree), kN Tensioner stiffness (linear), kN/m 300 o 4 343 225 650 300 Material X-S5 modified steel, yield, MPa Density, kg/m 3 Production and injection tubing 00 (5 1 12 in.), m ID,m Riser joint and coupling 00 (9% in.), m: ID,m Joint length, m 500 7850 0.1397 0.1243 .. Coupling mass added ttl .standard joint, kg 0.2445 0.2205 12.5 60.0 Fluid properties Annulus fiuid density, kg/m 3 Tubing fiuid density, kg/m 3 Riser coating and marine growth Rubber coating (lO-mm) density, kg/rna Everywhere Copper-nickel (1 mm) denSity, kgim 3 Above elevation of 268 n\ Marine growth (20-mrn) density. kg/m 3 Below elevation of 268 m HYdrodynamic parameters ... Drag coefficient . Below elevation of 280 m Above Inertia c o f f i c i ~ t Everywhere the bottom of the taper joint where fre- quency domain, 3-hour extreme (extrapolo- ation assumes a Rayleigh distribution) =481 kNm; time simulation, 30 minutes=772 kNm; and time domain, regular wave= 1044 kNm. It is evident at this location that the fre- quency domain results are seriously underes- timating the response. This is mainly caused by tension variations in the lower riser, where the effective tension is of the same order of magnitude as the tension increase resulting from the tensioner stiffness. Statistical extrapolation from the 30-min- ute simulation result to a 3-hour maximum based on a Rayleigh distribution gives a fac- tor of 1.16 on the 772-kNm value. The fac- tor found by comparison with the regular wave result is 1.35. This is an indication of the care necessary in using results from ir- regular time domain analysis. Fig. 2 shows the bending-moment enve- lopes for the three analysis methods. The above topics also are discussed later. The extrapolation method used for water particle velocity above mean water level is an important factor. The regular wave analy- sis result represents exponential extrapola- tionS (considered very conservative), whereas stretching method is used in the ir- regular analysis. S Frequency domain re- JPT March 1992 1025 10251600 1550 8300 0.5 ,0.5 1.0 2.Q suIts assume no wave loading above still- water level. Stroke Table 2 is an example of stroke results and use of load factors. The results are shown for 300-kN/m linear tensioner stiffness and top tension of 650 kN. The design load cases included unlimited limit state (ULS) = 100- year condition; progressive collapse limit state (PLSi)= IO,OOO-year condition, riser intact; and PLSdl = loo-year condition, one of four tensioner cylinders out of action. In addition to dynamic stroke, the follow- ing effects are significant: tolerances/ space- out, pressure range, and temperature range. For this case study a combined value of 0.5 m is included to account for these effects in the vertical-stroke requirement. For deck height and layout, such additional factors as Christmas tree height, tensioner size, neutral position, and installation of tools on the riser must be considered. Two approaches to determining stroke- range requirements are presented for this case study. First, if bottom-out and top-out are de- fined as events whose probability of occur- rence shall be less than 10 -4 per year, they may be disregarded in local design. The cor- responding stroke range for this case is 1.92 "Quantitative sensitivity results vary widely for the various responses along the riser, and the resulting priority is Influenced by those responses that govern the design." and is derived from load factors and analysis results shown in Table 2. Alternatively, a stroke range of 1.4 m may be adopted on the basis of a combination of serviceability and progressive collapse cri- teria. The serviceability criterion is that bot- tom-out and top-out are events whose prob ility of occurrence shall be less than 10- 2 per year (100-year event or ULS results in Table 2 without factors). The additional 10- cal design criterion to prevent progressive collapse in 10 -4 events is then a require- ment for an energy absorption capacity on the order of 0.25 MJ rather than the provi- sion of the additional 0.5 m required above. Fatigue AnalYSiS, Frequency Domain Fatigue analysis is a critical part of riser de- sign ,and will set design requirements to top tension, stress concentration factor (SCF) in components, marine growth control, prein- stallation inspection (quality control), and material properties. The target fatigue life is based on an assumed service life and a design coefficient of 10. The factor 10 is from Ref. 1 and cor- responds to the riser being classified as hav- ing major importance to structural integrity and having no access for inspection. An ad- ditional factor may be required, depending 327 Produclion Tre. _________ . OTC INJECT. RISER,T=650 kN, ------- " Of Is etc 20m, 100 yeor conditio.,45 deg I I I I I I )50 TLP Tr D ..... --------'==tR hf='---- TLP Service D.c" --==:C::J f- U ; r
l- I/ -
-
'" 2 .. c: 2,. 0 .. ,.
zoo - - .... RI r CouplI". ------.... e 150 f- - B ................. u ...... J ,,. - ....., hu . fI .... JI _h - ....-..... ,..t.,,. ..... J ,,, 100 T._ Joift. SO I- - Sub ... WOII ...... .- A o I _-._0 --- I noo 1000 lOG 100 400 zoo 0 - 100 Bending lIoment in KNm Fig. 1-TLP riser system schematic. Fig. 2-Bendlng moment envelopes for various analysis methods. on the level of corrosion protection nec- essary. In this case study an SCF of 2.5 is as- sumed for couplings even though values under 2 can be achieved. An SCF of 1.3 is used for reference should a weld-on cou- pling be chosen. Standard S-N curves (a plot of stress range vs. number of cycles to failure) as given in Ref. 10 are used ('B' for parent metal in the coupling and 'c' for the weld; no cut-off/threshold is assumed). The results for this case study are summarized in Table 3. Fatigue components caused by first-order wave response, low-frequency TLP motion, and vortex shedding induced vibrations are estimated. Table 3 is based on analysis using all sea states in the scatter diagram, a cosine 4 wave spreading, and eight wave directions. One linearization sea state (h s = 4.5 m, Tp = 10.5 seconds) is used to generate the stress transfer function at each location investigated along the riser. Upstroke\ ni 0;25 Downstrokil,Jl1 -:-0.65 toadfactor 1.3' ExPnslOn and;.spac&-OUti m _ .. 1 x 3 + 4 m " 2x 3= downstroke, m-O.85 Design strokli!,range, m ... O.83 to -1.09=1:92 TABLE 3-FATIGUE RESlJL TS (LINEARIZATION SEA STATE. hs =4.5 m, Tp =10.5 seconds) 'attgue Damagtnn fS --; ;.\ . at$d Esti .; . ,. '; " Estim m. . Elitimated .. 1 l\!litlmated . oamage, .. LocalionConeept Damage,_ 'pamage, " Vortex. Life .. ."'.A ... 'o..,;ng:::.,......R-'"is... er ___ {SCF{Curve) rlf$t .Ordef! seoond'1ill'der ! (Yta,f:s) iTaperbottom. 1.3JB "0:01610.0024,. ([0008 780; Mldwater threaded -c<lupling Midwater c<lupling .weld Splash-lOne threaded coupling Splash-zone ; coupling weld 328 2.SlB 1.31C 2.5JB 1.81C
0.0054 ... 0:0032 0.0000 0.0021'< 960-
0.0000 0.0008
2,541) 0.0000 0.0014 l.83Ir .
"I 0J)003 i 4.350' Table 4 shows results from a different linearization sea state (h s =6.8 m, Tp = 13.0 seconds). Comparisons of Tables 3 and 4 show that the higher linearization sea state introduces more damping, which, in the lower part of the riser, results in increased fatigue life. In the splash zone, drag is an excitation force and the fatigue life is re- duced correspondingly in this area for linearization in a higher sea state. The choice of linearization sea state, therefore, is important; and, as shown by comparison of Fig. 3 (distribution of h s ) and Fig. 4 (damage distribution of h s ), the proper approach is to use multiple sea-state linearization. A typical North Sea scatter di- agram has around 100 sea states with sig- nificant levels of occurrence. Linearization in each sea state is possible, but around 10 representative linearization sea states pro- vide reasonable accuracy. The choice of sea states and appropriate "windowing" of the wave scatter diagram is important and should include results along the riser. Anal- ysis with many windows can be used to calibrate simplified methods during design development. Fatigue Analysis, Time Domain Variations in axial tension occur in a TLP rigid riser and have been considered for par- ticular sea states using time domain analysis. Tension variations lead to axial stress cycles, but there also can be a significant variation in the cable stiffness of the riser during the passage of a wave. To quantify the influence of this, damage in a 3-hour sea state is solved with the fol- lowing methods: (1) frequency domain anal- ysis and assuming a Rayleigh distribution March 1992 JPT F AT I GUE DAMAGE va. SIGN. VA VEI- I GHT SIGN. WAVEHEIGHT PROB.DISTR. o 01 ) (f)
U Z
ex: :::> u u 0 o "- CD !!! ri >- 0 o 0 Z
4.1, 1.1 "'lINEAAISATOI N1 !:! z ... 5, 6 5 LlNEARISATION ex: .Ii SASTATE o SEASTATE w m 8.5 to.' 1: 10.5 12.5 0
o Taper-bottom Midwater Fig. 3-Design field significant wave height distribution. Fig. 4-Fatlgue damage as a function of significant wave height. of stress cycles, (2) irregular time domain analysis with variable tension and a rain- flow stress cycle counting method, 11 (3) as Method 2 and stress cycle counting assum- ing the stress peaks are mirrored about the mean, and (4) as Method 2 and stress cycle counting assuming the negative stress peaks are mirrored about the mean. Significant differences in results occur be- cause of (1) different resulting standard deviations in time domain vs. frequency do- main (Tables 5 and 6), (2) different zero- crossing periods (Tables 5 and 6), (3) non- symmetry of the response (Figs. 5 and 6), and (4) non-Rayleigh probability distribu- tion of stress cycles (Figs. 5 and 6). Table 5 shows significant differences un- der Items 1 and 2 above when comparing time vs. frequency domain. Item 3 is also seen in Figs. 5 and 6 to be a significant fac- tor contributing to this difference. The in- fluence of Item 4 is less significant except in the splash zone, as seen by comparing re- sults from Methods 2 through 4. Table 6 is for a higher sea state where hs = 6.5 and Tp = 13.0 seconds. In this case Items 1 and 2 are in good agreement. How- ever, the resulting damage is reduced in the time domain in this case, thus reversing the trend in Table 4. The comparisons reported here are incon- clusive. However, the main point is to high- light the possibility of using time domain methods for calibration of frequency domain results. Time domain is not a practical de- sign tool for fatigue analysis because of computer-time requirements. It is interesting to note in the time domain that rain-flow cycle counting is close to Methods 3 and 4 away from the splash zone and consistently gives less fatigue damage. "Negative stress peaks symmetric" (Method 4) is consistently worse. Interference Interference is an important factor to con- sider for riser spacing and layout. Inter- ference places restrictions on dynamic similarity between risers. The drilling riser JPT March 1992 and the production/injection risers should have similar natural frequencies to minimize relative motion. Results for intact risers, therefore, are sensitive to top tension. Spacings of 3 to 5 m are feasible depend- ing on the concept and location; however, margins to export risers should be greater to minimize progressive damage potential in the event of tensioner failure. The spacing to the pontoons also should be greater. Weather limitations for riser running also is an important consideration and is much improved with increased spacing. Frequency domain analysis is used to car- ry out interference analysis. The critiera are no impact in the lOO-year storm (i.e., most probable maximum relative motion in 3 hours does not allow contact) and no pro- gressive collapse for the maximum impact velocity solved in PLSi or PLSdl conditions as defined previously. "The choice of sea states and appropriate 'windowing' of the wave scatter diagram Is important and should Include results along the riser." Should impact occur, post-impact analysis is necessary to evaluate progressive col- lapse. The relative motion of the risers is assumed to be a Gaussian process and the impact velocity is estimated from V max =u v .J2 In N-(Z/uZ)2, ..... (1) where V max = most probable maximum im- pact velocity, u v = standard deviation of TABLE 4-FATIGUE RESULTS (LINEARIZATION SEA STATE. hs =6.8 m, Tp = 13.0 seconds) . Patfgue Damage in 15 Years . Estimated Estimated Estimated Damage, location Concept Damage, Damage, Vortex Along Riser . (scflCurve) First Order Second Order Shedding Taper bottom 1.3/6 0.0086 0.0024 0.0008 Midwater threaded coupling 2.5/B 0.0084 0,0000 0.0002 Mldwater coupling weld .1.3IC 0.0036 0.0000 0.0006 Splash-zone threaded coupling 2.51B 0.0114 0.0000 0.0014 Splash-zone cou- piing weld 1.3/C 0.0047 0.0000 0.0003 Estimated Fatigue Life {Years) 1,270 1,435 3,615 1,180 3,000 329 Taper-bottom Bend.Strs UPa Global Maxima 30 min Toper-bottom Bend.Strs MPo Clobol Minima 30 min ToI.Obs.23I,Hsa . 5111.T,aIO.5s _ :: r-----"r------r=r-::---..... ----r-----r-,----. c: ... .. "" ... - ... .Q ... .. .0 .. .. ... . CL ... .. ... ... ... ... - ..... '.5 .. C> .. "'a.0 '.5 1.0 1.5 2.0 2.5 3.0 1.' 1.5 2.0 2.5 3.' Number Of Standard Oeviotio., Number Of Standard Deyiations Fig. 5-Stress peak distribution at taper bottom, hs =4.5 m, T p = 10.5 seconds. Fig. 6-Stress negative peak distribution at taper bottom, h s = 6.8 m, T p = 13.0 seconds. relative velocity, N = number of zero cross- ings for the relative velocity in 3 hours, Z=mean separation in interference analysis, and az=standard deviation of relative dis- placement. sign is very sensitive to tensioner-stiffness characteristics; in addition, the worst-case loading varies along the component. There- fore, this is a component that requires much optimization effort and restrictions will be placed on characteristics of other compo- nents in doing so (e.g., tensioner stiffness). "Should impact occur, postimpact analysis is necessary to evaluate progressive collapse." Components and Interfaces Interface data are critical for design of the deck and template/subsea wellhead and, therefore, are very important. The subsea interface data may be required up to 3 years before the riser is installed on projects where predrilling is maximized and, therefore, should include an appropriate contingency. Riser coupling types for this size include integrated threaded, weld-on threaded, and bolted flanges. All options are feasible, and preferences will depend on specific project require- ments. Integrated threaded couplings are ec- onomically attractive; however, fabrication limits on the thickness and length of the up- set can make this solution marginal in some applications. Detail design considerations for threaded couplings are given in Ref. 12. Flanges should be of a special design (e.g., compact flange type) optimized for this ap- plication. Taper bottom Midwater Splash zone Taper bottom Midwater Splash zone 330 A taper joint is included in the production! injection riser to minimize the stress ampli- tudes in the production tubing. The length is limited by handling and fabrication con- siderations, with a reasonable limit around 18 m. The profile requires design iterations con- sidering both fatigue and extremes. The de- Ref. 3 is recommended for further details of components and for local design criteria. TABLE 5-FREQUENCY/TIME DOMAIN FATIGUE COMPARISION, hs = 4.5 m, T p = 10.5 seconds DAMAGE DURING 3-HOUR EXPOSURE (1.0/B) FACTORED BY 1 x 10 8 Frequency Domain Standard Deviation Zero Crossing Bend (MPa) Period 8.72 7.93 5.00 7.29 4.02 6.78 Damage Rayleigh 76.1 11.6 5.2 Time Domain Damage Calculation Method Standard Deviation tero Crossing Rain-Flow Peak Positive Peak Negative Bend (MPa) Period Counting Symmetric Symmetric 9.34 7.80 91.8 97.1 100.3 6.47 7.35 25.9 25.6 26.4 4.54 6.41 10.2 8.0 15.6 TABLE 6-FREQUENCY/TIME DOMAIN FATIGUE COM PARISI ON, hs = 6.8 m, T p = 13.0 seconds DAMAGE DURING 3-HOUR EXPOSURE (1.0/B) FACTORED BY 1 x 10 8 Frequency Domain Standard Deviation Zero Bend (MPa) Period 11.58 9.67 5.41 7.89 5.48 7.58 Damage Rayleigh 251.2 14.7 16.1 Time' Domain 'Damage Calculation Method Standard Deviation Zero Crossing Rain-FIOV/ Peak Positive Peak Negative Bend (MPa) Period Counting Symmetric Symmetric 11.64 9.53 166.2 169.6 182.4 5.61 8.08 13.3 14.4 13.8. 5.52 7.04 22.8 .. March 1992 JPT Conclusions 1. Riser interfaces are fabricated at an ear- ly stage in a TLP project, so early progress of the rigid riser design requires high pri- ority to maintain economic options. 2. Time domain analysis is necessary for analysis of extreme events. 3. Progressive collapse criteria signifi- cantly affect the design. 4. Simplified analysis methods optimize the number of design cycles. Maximization of design cycles is important. 5. A TLP rigid riser system is sensitive to many parameters, and, in addition, such components as the tensioner and taper joint are interactive. 6. Allowances for deviations from base- case-riser assumptions should be maxi- mized, as such will occur during detail de- sign of components. Nomenclature h s = significant wave height Tp = wave spectrum peak. period References 1. "Regulation for Structural Design of Load- Bearing Structures Intended for Exploitation of Petroleum Resources," Norwegian Petro- leum Directorate (Oct. 1984). 2. "Draft Guidelines for Design, Fabrication and Installation of Submarine Pipelines and Risers," Norwegian Petroleum Directorate (1988). 3. Alstad, O. eta!.: "SnorreRiserand Well Sys- tems," paperOTC 6624 presented at the 1991 Offshore Technology Conference, Houston, May 6-9. 4. FREECOM Program Manuals, Marine Com- putation Services Inti., Galway, Ireland (1988). 5. FLEXCOM Program Manuals, Marine Com- putation Services Inti. , Galway, Ireland (1988). JPT March 1992 6. Martinsen, T.: "Hydrodynamics in TLP De- sign," OMAE (1989). 7. "Rules for the Design, Construction and In- spection of Submarine Pipelines and Pipeline Risers," Det norske Veritas (1981) 64. 8. Rooney, P., Lereim, J., and Madsen, H.O.: "Application of Probabilistic Methods for Verification and Calibration of a TLP Tether System Design Based on Partial Coeffi- cients," OMAE (1989). 9. Sparks, e.S.: "The Influence of Tension, Pressure and Weight on Pipe and Riser Defor- mation and Stresses," ASME (1984) 106. 10. "Fatigue Strength Analysis of Mobile Off- shore Units," Classification Note No. 30.2, Det norske Veritas (Aug. 1984) 64. 11. de Jonge, J.B.: "The Monitoring of Fatigue Loads," NatI. Aerospace Laboratory (NLR MPOO7U) , The Netherlands, ICAS-paper 70-71. 12. Kirkemo, F. and Hessen, G.: "Structural De- sign of Threaded Couplings for Potential Use on Rigid Risers," OMAE (1990). 51 Metric Conversion Factors bar X LO- ft X 3.048- Ibm/U.S. gal x 1.198264 quad x 1.055 056 psi x 6.894 757 U.S. tonf x 9.964 016 U.S. tonf-ft x 2.711636 'Conversion factor is exact. Provenance E-Ol = MPa E-Ol = m E+02 = Kg/m' E+12 = MJ E+OO = kPa E+OO = kN E+OO = kNm Original SPE manuscript, TLP Rigid Riser: A Case Study, received for review May 7, 1990. Revised manuscript received Jan. 3, 1992. Paper accepted for publication Jan. 8, 1992. Paper (SPE 20869) first presented at the 1990 Offshore Technology Confer- ence held in Houston, May 7-10. JPT Rooney Engebretsen Phe"m p. Rooney holds a BE degree from Unlve ... lty C. in Galway,lreland, and an MS degree lrom Cranfield In$t. 01 TechnolOgy, U.K. He was seconded to Saga Petroleum AJS, has worked with the Snorrerlgld rise,. $Ince 1988, and recently begalil workllil9 at Conoco Nor- way.Knllt e .......... t .. nholds aBS degree In mechanICal elil9lneerllil9 from of Utah and MS degree$ln me- chanICal and ocean e:nglneerllil9 from the MusachuMtts lnat. of Technology. He: is discipline supervisor for Troll Phase 2, Basic Englneer11i19 for SUb- atructure(Concrete Floater) atNor. weglan COntracto .... Aker Engineering. Dee ""'ell .ett-:een earned an MS degree from the Norwegian Inet. of Trondhelm. At Saga Petro- leu", he Is responsible for suCh are&$ .. global TLP enalysls,rlHr analy8l$, and modelltestllil9forthe Snorre ProJect: unavailable.) . 331