a a AR ae a A i eer A mer P. Papalambros Assis, Profesor, Dept. of Mectanical Engineering and Applied Mechanis, M, M. Bernitsas Assis, Professor, Dept of Naval rehitecure ‘nd Marine Engineering The University of Michigan Ann Asbor, Mich. 48108 Introduction ‘The marine riser is an extension of the well at the ocean bed. It is a long tubular structure connecting the well head with the offshore platform. The riser guides and protects the dlril string and circulates the drilling mud. The first riser was installed in 1949 at a 20 ft depth. At present, 4,000-ft risers arein operation and 7,000-(t risers are inthe design stage. The static and dynamic riser behavior has been studied by various authors numerically (1, 2] or analytically [3, 4]. However, several problems associated with riser mechanics have not been solved as yet and substantial theoretical work is still needed [5-7]. The first attempt to define the feasibility domain and solve the optimization problem was done by the authors of this paper (3]. The problem of the riser response under generalized static load was formulated and solved analytically [3] in terms of Airy functions of the first and second kind. The hydrodynamic loads were modeled in terms (of a general polynomial expression of the water depth. This is mandatory for a rigorous riser design since at the current state of the art the knowledge on the hydrodynamic loads is limited, In the present work analytic approximations to the exact solution are derived and used to express the constraints and objective in the optimization problem. The advantage of analytical solutions isthe fact that they reveal the dependence of the riser behavior on the design variables, parameters, and It is desirable to find at least some analytical expressions relating the optimal values of the design variables with the riser parameters. Moreover, it is important to be able to recognize which design requirements are critical at the op- ‘imum. The model itself is not readily amenable to numerical ‘Weatment, particularly if results useful in design practice are to be obtained. Thus, monotonicity analysis (8-10) is an appropriate method for a preliminary investigation of the ‘optimum. Identification of active constraints prior 10 the ‘numerical implementation reduces the degrees of freedom and ‘makes the location of the global optimum possible as well as rigorous {8}. The overall optimization procedure becomes ‘more efficient and the results become more meaningful for the designer. | Conte by the Design Automation Commitee for publication inthe Joust oF Mechanica, Deak. Manuscript ecived at ASME Henequares Sane 19 Journal of Mechanical Design Monotonicity Analysis in Optimum Design of Marine Risers The optimal design of marine risers used for drilling and production of oil in off- shore operations 1s studied. The optimization problem is formulated on a two- dimensional model for bending of circular tubular beams’ under tension and In- ternal and external siatic pressure. A general polynomial expression describes the external hydrodynamic toads. Monotonicity analysis is used 10 identify active constraints, determine design rules, and reduce the size of the problem. wearized Static Riser Model ‘The general nonlinear, large deflection, three-timensionat model, derived in (5], which describes the dynamic behavior of a riser including its extensional oscillations, can be sim- plified to yield the model used in the present feasibility and optimization design analysis. The linearized differential equations of equilibrium of a riser subject to small deflections and slopes in a vertical plane Equilibrium of moments (see Fig. 1) aM _, a “ie PiGe +O=0 o Equilibrium of forces inthe x direction 4 a @ Equilibrium of forces in the zdireetion dP. mit ® where Ws Wat Wat We—B ® =D," aD? pete. a Teste thet) —ta thet) 1) U isthe deflection of the riser in the x direction, T(z) isthe actual tension inthe riser, M@) isthe bending moment in the y direction, Q(z) istheshear force in the x direction, J.(z)_ is the external force per unit length exerted on the Inthis problem f, is considered time invariant. Inaddition, * (D3 -DAI 6) we] (Det ~DeY c is the weight of the riser per unit length, o js the weight of the drilling mud per unit length, OCTOBER 1982, Vol. 104/848(8a) vw (Ds? ~D,?) is the buoyancy of the buoyancy modules per unit length and Baa, ED? (80) is the buoyancy of riser and modules per unit length. ‘The linearized constitutive relation of bending is eu Mc) =e ° where 80. a0 Finally the boundary conditions are: ay (2) a3) ay The static external hydrodynamic force, f,(2), is propor tional 10 the square of the relative fluid velocity. Along the Fiser (depthwise) we can approximate the force profile by a rth degree polynomial as in equation (15), $= $9-6oVDy [ape (+) +02(2)* + ean +0(Z)'] where W isa characteristic velocity and Cy is the drag coef- ficient. Equations (1), (2), 3), and (8) can be combined t0 vield the fourth order differential oquation with variable Covfficients which describes the risers response to external Static loads: @ fp, @U)_¢ 5 & [SZ ]-2 [oneresoy AA better understanding of some of the terms involved in evations (1) 0 (16) i requized for proper formulation ofthe ‘optimization problem. Specifically, W(z)the effective weight of the riser per unit length and F(z) the effective + as) f(z) auc) 16 Nomenclature z Rergan Meerg.de kenge) (yee Queda Wiferde Me-jaz> ace-gazs he ae Pa Fig. 1 model Free body diagram for #ailteren lament de; nearzed tension defined by equations (4) and (5) are two key variables of the design problem [5]. The effective weight of the riser per unit length, W, (2), is equal to the weight of the riser WV, plus the weight of the ‘mu, plus the weight of the buoyancy modules W/,, minus the total buoyancy B per unit length. That is, H(z) is the weight of riser and contents in water per unit length. In equation (3) describing the force equilibrium in the 2 direction it appears as if the force exerted on the differential element dz (Fig. 1) is equal to H(z) edz. This is not correct since dz is not a closed body fully wetted by water. For this reason the left-hand side of equation (3) is not the actual tension T(z) in the riser but Pe(z). P,(z) is the effective tension, i.e. the actual tension T(z) modified due to the internal mud static pressure and the external hydrostatic pressure [3]. In this paper W/.(z) is considered constant and con- sequently P,(z) is a linear function of 2. As a result, the design parameters in equation (10) are W, and P,(0). This conclusion is particularly important in the derivation of the buckling constraint. The buckling loads are functions of the effective tension at the riser's lower end P,(Q) (7] and not 7{0). In reference (7) it is proved that a tiser may actually buckle as a Euler column even if the actual tension in the riser (2) is positive forall values of z between O and L.. average diameter of riser ‘outer diameter of buoyancy modules, maximum allowable Dy inner and outer riser diameters minimum allowable D, Young's Modulus Tiser cross-sectional area moment of inertia riser’s length bending moment effective tension shear force constraint no. i riser's material yield strength actual tension at the riser's lower end iension at the top ofthe riser riser's lateral displacement effective weight of riser per unit length weight of drilling mud per unit length 850 Vol. 104, OCTOBER 1982 weight of riser per unit length total height of mud column total water depth external hydrodynamic load per unit lengtiy dimensionless vertical coordinate along. the Position of maximum total stress ensionless effective riser weight specific weight of buoyancy modules; materia, riser wall material, water specific Weight of circulating mud lower and upper limits for Ym static offset of drilling vessel principal stresses inr,r,z directions bending and effective stresses ‘combined Tresca stress 0. Equality Constraints. The following equality constraints are derived from the relations developed in the riser model section, Equations (4), (6), (7), (Ba), (8b), (10) and (18) yield: E/B/L) = We + (wm/4)D2 —(44/4)Da? sr model, + (#yn/4(Dg? Dy?) () Wy= (my9/4\D,? -D?) oy 1/64N.D,* ~D,*) (62) Elimination of P,(z) from (3) and (5) yields an expression for T(z). Forz = £, T(z) is equal tothe tension atthe top of the riser (TTR) given by TTR = Wel +10) ~(mywa/4ND. =De)L (63) Expressing (63) in terms of the nondimensional variables vields (HEIL =TTR+ (4 Y4L/Dy~D,?) (64) Practical Constraints. Drilling mud circulates in the drill sicing and between the drill string and the riser. Its function is, to facilitaw drilling, carry away cuttings and cool the drill, string. For these reasons a minimum mud circulation is Fequired which depends primarily on the drilling depth. The foregoing requirements impose the following constraint on the internal riser diameter. RED 2D 65) ‘The density of the drilling mud ym depends primarily on the drilling depth below the sea bed and also on other factors like lubrication, protection of riser from the drill siring, and prevention of blow out, These definea rather small acceptable range of values for yy, expressed by constraints R,, and Rooks?) Sm Sng (66) To keep the riser up, prevent buckling or collapse, and provide additional strength to lateral loads, a tension is ap- plied at the top of the riser (77R). This force, provided by the tensioning system installed in the drilling vessel, is very high patticularly for long risers. To alleviate high tensile stresses imposed at the upper end, buoyancy modules are distributed along the riser thus reducing the effective riser weight in water. Part of the required upward force is provided by the tensioning system implying the constraint Ry:TTR>O © Additional upward force is provided by the buoyancy ‘modules. This partial buoyancy implies the constraint Ry: Wp — (ry w9/4MDy? ~D,2)20 (68) where yun = yw—a. The equality holds for a neutrally buoyed "riser. “The "buoyancy modules increase the hydrodynamic loads exerted on the riser. To avoid excessive ‘external loads and high constcuction costs the size of the buoyancy modules is limited by oo) Journal of Mechanical Design Finally note that all design variables are strictly positive. An obvious geometric constraint is Ry:Dp>Dy 70) considered inactive in advance, because only a design con- figuration with buoyaney modules is examined. Monotonicity Analysis Monotonicity rules will be used to identify inequality constraints that are active at the optimum (8, 10]. To apply these rules the problem is cast into normalized form and equality constraints are eliminated from the model along with aan equal number of design variables. Furthermore, in order to expose monotonic structure of the functions involved the following transformations of variables are used. a) (72) D, + 2, whete the thickness ofthe riser tube ‘is very small compared to either D, of D,. Thus, Dy? >> ‘Dg? with a maximum error of $ percent and this result i used to: simplify. some expressions. ‘The sesulting optimization problem in normalized form isthe following. Problem Pl min Wy= (ern /4)D a? subject 10 g, CUENTA PB. 8 CSTW ID.Dye (S/N NOL? GeV LDAP) (S,/N/ De? 2 £6D,7(»0+ 5) Sn (8202 /V7a) PB.) p Ru Cy (SIN Dade LDA (S/N PDe yw ILD, (S/N we Rut Diag 7S 1 Rst Da Don S1 Res my tm PSI Rit my "Yn SI Rat UDg°/D,)~ Aral ya) (Da? /Ds SA 2D _ Yon 28 2 1 Da? _ EBDg? Be Dy DE Bal” 2 DDD? <1 2D,/D,2)- 8+ BEDE aL? <1 1BDu Da Davin >0- Note that Ry is derived from (60) by elimination of Zand We. Also, constraints Ryo, Ry) are assumed inactive, i.e. only the ‘ease with buoyancy madules and a riser under tension will be examined, OCTOBER 1982, Vol. 104/853Consider the variable + which appears only in constraints R, and R; and notin the objective. Then, either both Ry and Ry are active, or they are both inactive, In fact, if either of them were inactive, the other would have to be also inactive, since otherwise 7 would appear in a single constraint and its value at the optimum would be determined directly from the optimal values of the remaining variables. This srgument holds whenever the single constraint does not imply another ‘modified constraint on the remaining variables, as for ‘example in the case of asymptotic substitution (10). Next, assume that Ry and Ry are both inactive, Constraint Ry shows that Dy decreases with respect to. Implicit elimination of Dy from the model with the aid of Ry results to Problem P2 min Wy (Dy*) subject 10 Ralra* Da" De-VS1— Rolin) S1 RWDA-) 0). Tius the case of both Ry and Rp being active is, described by Problem P3 min We (De*) RiB.7Du* Dae RIG, nDe* Dy Dy RMDa+.Da7 <1 RAD VSI RelDa" 1 Rs(G,Ds,Ds,Dg)=1 This problem cannot be reduced further unless the external loads are specified in order to obtain exact relations in Ry and Rs For Py (8.7), Pr (Ber) Py atid Bo 1854 / Vol, 104, OCTOBER 1982 Discussion ‘The original optimization problem was reduced to problem PI with five degrees of Ireedom and eleven constraints. This was further reduced to problems P2 and P3. Problem P2 yields @ consiraint-bound solution in closed form which is a local optimum for problem Pl. Problem P3 having wo degrees of freedom can be solved numerically once the design parameters and constants for a particular application are known. Any of the local optima obtained should not violate ‘any of the constraints assumed inactive. From the design point of view it is found that the mud density must be set always at its lower limit. In addition, at least one siress constraint must be active. In particular, when the internal loads are predominant, ie. when the mud static pressure induces higher loads than those induced by the ex- ternal hydrodynamic forces and the tension along the riser, constraint Ry isactive (problem P2). A closed form solution thus obtained which does not depend on the external loads. Furthermore, when the external loads become significant, the ‘combined stresses both at the top and the bottom of the riser will be critical (problem P3), Conclusions ‘The original exact solution to the riser model was adequately approximated so that explicit expressions for the stresses were obtained. This allowed an algebraic formulation of the minimum weight design problem. Subsequent ap- plication of monotonicity analysis provided one local op- imum and desian rules for the investigation of other possible local optima. Numerical solutions can be obtained in future research following the procedure described in this paper and establishing the values of riser parameters, and the expression of the hydrodynamic loads. Parametric analysis can then generate design charts for practical use Acknowledgments The work of the first author has been supported by NSF Grant No. CME 80-06687. Both authors have been supported by Faculty Research Grants No, 387569 and No. 387565, respectively, from the Horace H. Rackham School of Graduate Studies at The University of Michigan. This support is gratefully acknowledged. References 1 Azar, J. Jo and Solve, RL E., A Compechemive Study of Marine Dring Rivers," ASHE Paper Peon Divison, PET, 1978 2 Sparks C. P., "Mechanical Bebrior of Marine Riser Mode of In ‘Maen of Pricial Parameters, Dezpwater Mecriag and Driling.” A SALE OD. VoL? 1918 4 Bernas NM. snd Papalambos,P, "Design Opinion of Risers Under Generlze’ Stic ose Inermartec’ 80 tre. Sep 108 “tHapel KH, "Siucsueal Anais for 3 Long Diling Rise ‘erocean 6. tn German) hernias, MM Mode of Byram Responeof Risers, Pipelines Resear, 12. “ Berblsas 1. NE, “On the Design of Marne Rises,” S.NLA.ALE: New Enplend Section, May 1979. ST Bernas MMM, “er Top Tension and Riser cling Lond, ASAIE Comptia! Sterna for Offthore Serctues, AMDNNol 37, Nov 8 Wie, D. ., Globaly Oprmal est, Wiley tmersienes, New York, we PapalanbrosP. Thyce Dimensional, Nonlinest Large Deestion Cab, Journ! of Ship _"Monekonii n Goa and Gromtric Progamming. ASME Joust or Mecinstat Dest, Vol 4, Nov Jan 983 “Glob Nonterative Deion Op ASME Jovnsat oF Mecassten, 10 Papatumeo, Pan Wile, Do. simiation Using MowsioniciyAnabsi Denn, Volt, Nov 8; Oa. 70