Spectral fatigue for FPSO conversion.

Transcription

1 Spectral fatigue for FPSO conversion. Vincent Bonniol, Introduction The required time to build a new FPSO leads to more and more conversions from existing tankers. On demand of several oil companies has carried out during the past years a large number of spectral fatigue analyses to assess the behavior of the hulls on sites everywhere around the globe. Until a few years ago the main method for fatigue assessment was the well known deterministic methodology for FPSO S, adapted from the Rules deterministic fatigue for trading Tankers. The spectral fatigue is a different approach, where the effect of the wave on the structure is considered linear. The motion analysis is performed in order to estimate the wave-induced loads to be considered in the SF analysis, including external wave pressures, internal tank pressures due to tank liquid accelerations and inertial force on the masses of significant equipment items. Then, structural analysis is performed in order to generate the transfer functions for both Tanker and FPSO phases. Separate 3D finite element global models are built for the tanker (accounting for her asbuilt configuration) and for the FPSO, incorporating the project structural modifications, loads and scantling diminution due to corrosion based on latest thickness measurements. Several local models are built for both Tanker and FPSO configurations in order to generate the stress transfer functions used in the spectral analysis for hundreds of hot spot locations analyzed for the different drafts, headings and frequencies. Once stress transfer functions were available, spectral analysis are carried on for each of the hot spot elements above mentioned. Wave data are incorporated to produce the stress response spectra to derive the magnitude and frequency of local stress ranges for each element. For the Tanker phase the wave scatter diagram is derived from the vessel trading history and associated navigations zones based on the Global Wave Statistic database. For the FPSO phase, the site specific conditions are taken into account and metocean data was based on Heading Analysis study. For each calculation some non-linear correction such as wave spreading, rainflow and intermittent wetting are applied on the process. Finally the in progress developments will be described, such a new SF methods based on 3 holds models, decreasing significantly the modeling and the running time.

2 Deterministic method: The deterministic method or one wave method is directly derived from the traditional method used for tanker according to the BV Rules (see BV Rules for steel ships NR 467 and Offshore Rules NR 445). A hydrodynamic calculation is performed in order to obtain the load parameters for several different drafts (at least minimum, maximum and one intermediate draughts are required) on the main wave loads: - vertical and horizontal bending moments - Shear force - Relative Wave Elevation - Different accelerations along the ship The long term distribution for these items is obtained and the values at the required wave return period may be applied on the calculation model. It has to be noticed that here is the difference with the tanker Rules approach, where the loads parameters are not determined through hydrodynamic calculation but are described in the Rules. The stress ranges are obtained through the FE calculations of crest and trough loading cases. Knowing the stress range at one probability and the long term distribution the damage for each considered pattern is obtained through the Weibull approach. This means that the long term distribution curves are shaped by a Weibull coefficient. LOADS Long term distribution DAMAGES One wave loads and stresses Figure 1: Deterministic fatigue

3 The Spectral fatigue analysis: Presentation Unlike the deterministic method, in the spectral fatigue procedure the hydrodynamic analysis is used for applying pressures from calculation along the structural model. The methodology is now to determine, through the BV hydrodynamic software Hydrostar those pressure for a set of drafts, directions and frequencies. Then the site meteocean data will be used for obtaining RAO and final damage calculation. LOADS Stresses (frequency damage) DAMAGES Site meteocean data Scatter diagram Figure 2: Spectral fatigue In order to perform the SF method described here above the following steps are carried out: 3D-difraction radiation analysis based on global hydrodynamic model to derive dynamic pressures and accelerations for each sea state for each loading condition. 3D global analysis based on selected loading conditions and dynamic loads derived from hydrodynamic analysis for each load case defined by each internal loading condition, sea state and heading. 3D local structure analysis for selected areas accounting for loads and boundary conditions generated from the global analysis to obtain stress response due to each sea state and for each loading condition. Fatigue damage computation for each loading condition according to the combination of heading/frequency corresponding to site conditions (for FPSO), trading routes (for tanker) and internal loading (for FPSO loading/offloading sequence). Fatigue strength assessment for each analyzed detail by calculation of total cumulate fatigue damage and estimate of remaining fatigue life.

4 The FPSO phase is not the only one to be analyzed with spectral analysis. It is crucial for the validity of the final results to account correctly the damage accumulated during the tanker life of the vessel. It is a common assumption to consider that ocean waves are the main source of the fatigueinducing loads acting on the structural system being analyzed. The fatigue damage from other loading sources might be considered separately. However for some special cases the rainflow method may be applied to account correctly the combined stresses and cycles. In order for the frequency domain formulation and the associated probabilistically-based analysis to be valid, load analysis and the associated structural analysis are assumed to be linear. Hence, scaling and superposition of stress transfer functions from unit amplitude waves are considered valid. Nonlinearities, brought about by nonlinear roll motions and intermittent application of loads such as wetting of the side shell in the splash zone, are treated by correction factors.

5 Motion analysis The determination of Response Amplitude Operators (RAOs) is the final target of the hydrodynamic analysis. In addition to vessel motions, the other wave-induced load effects considered in the Spectralbased Fatigue Analysis are mainly the external wave pressures, internal tank pressures due to tank fluid accelerations and inertial forces on the masses of structural components and significant items of equipment, as deckhouse, topsides, helideck and special offshore items. For the tanker phase, calculations are performed based on a speed equal to 75% of the maximum speed of the vessel. The total hydrodynamic pressure includes the direct pressure components due to waves and components due to hull motions. The components of the hydrodynamic pressure have been determined by means of 3D diffraction-radiation analysis using program HydroStar. In order to generate the stress transfer function for both Tanker and FPSO phases, thousands of load cases are generated as follows: 2 tanker loading conditions: Full and ballast 3 or 4 FPSO loading conditions. Basically the max and min drafts will be selected. In addition 2 intermediate cases may be chosen. In order to catch the maximum stress ranges applied on the structure they may be alternate conditions, such as chessboard patterns for example. A large number of heading is required. 12 for instance, (every 30 between 0 and 360 ) Several frequencies are necessary to obtain acceptable RAO s. 25 frequencies, from 0.1 to 2 or even 6rad/s are a typical choice. Real and imaginary cases Figure 3: Hydrodynamic model

6 3D global analysis The global model is the first step in the finite element calculation. It is built in order to represent the behavior of the primary structure. The full ship is modeled, from bow to stern, including any item which might modify the weight distribution along the vessel. Dynamic pressures and accelerations for each combined load case obtained from 3D diffraction radiation analysis above described are transferred to respective Global 3D FEM models and structural analysis are performed using program VeriSTAR CSM. Figure 4: Structural models

7 The tanker models (both global and local meshes) are modeled as built, according to original tanker design and may account corrosion margins. Methodologies such net approach for instance may be used. Regarding FPSO models the last gauging report is used to account the real scantling of the structure. Corrosion margins may be used. Modifications such additional cofferdams, oil tight bulkhead, turret moonpool and miscellaneous reinforcements are incorporated into the FPSO 3D FEM global model. Topsides are modeled up to the first main equipment deck level to ensure that the interfaces between topsides are properly modeled. Topside loads are applied at the modules center of gravity with considered load cases accelerations and transmitted into the pallet structure and supports. If items as flare stack and turret gantry are not entirely modeled, their loads have to be taken into account properly. Mooring effects might be considered not relevant for the analysis due to the fact that the stiffness and damping of the mooring system are second order motions of the ship that have much larger periods. Therefore, mooring forces may be modeled with concentrated loads.

8 3D Local analysis Refined meshes are then created in order to provide more accurate results for selected areas. The typical mesh size for specific fatigue fine mesh models is between 1 and 2 times the element thickness in the hot spot area. The load cases verified in the local fine mesh analyses are imported from the Global 3D FEM analysis. The displacements obtained from the global analysis are imposed as boundary conditions for the refined model and loads are also automatically imported. Modeling and calculations for local 3D FEM models have been also carried out using the program VeriSTAR CSM developed By. For each local model, two load cases corresponding to the real and imaginary parts of the frequency regime wave-induced load components are analyzed and for each heading angle and wave frequency, the frequency-dependent wave-induced cyclic stress transfer function, Hσ(ω θ), is obtained for the Base Vessel Loading Condition and vessel speed, as applicable. Figure 5: Very fine mesh model Side Shell Longitudinal

9 Figure 6: Toe of transverse bracket Figure 7: Topside analysis

10 RAO construction The RAO for each element and each loading condition is built according to the following procedure: The first step is to extract the centroidal values of the surfaces (upper and lower faces) stress components for the stillwater case and for the real and imaginary load cases. Interpolation methods may be used in order to obtain the stress, for example linear extrapolation over reference points 0.5 and 1.5 x plate thickness away from the hot spot. Otherwise the stress will be taken at 0.5 x thickness of the hot spot element, which is modeled t x t (thickness x thickness). These outputs allow us to calculate the dynamic part of each real stress components by subtracting out the still water case. The second step is to calculate the amplitude of each dynamic stress component Principal stresses 1 and 2 by taking the combination vector of the dynamic parts of the real and imaginary components. The largest principal stress amplitude will be used to define the first frequency component of the discrete stress transfer function. These 2 steps will be repeated for other frequencies and other headings to construct the complete discrete stress transfer function. Figure 8: RAO s for 12 headings and 25 frequencies

11 Damage calculation The wave data are incorporated to produce stress response spectral, which are used to derive the magnitude and frequency of occurrence of local stress ranges at the locations for which fatigue damage is to be calculated. Wave data are represented in terms of a wave scatter diagram and a wave energy spectrum. The wave scatter diagram consists of sea states, which are short-term descriptions of the sea in terms of joint probability of occurrence of a significant wave height, Hs, and a characteristic period. S-N curves and Miner s damage rule were used to calculate the fatigue damage cumulated during both tanker and FPSO phases. Cumulative fatigue damage is computed from the long term distribution according to Miner-Palmgren law: This means that we assume that the cumulative damage from the total amplitude stress cycles (the group of sea states) is similar to the sum of the damages inflicted by each single stress cycle (each sea state). The sequence in which the stress cycles occur is not accounted. The failure happens when the cumulative damage is equal to one. Computation of fatigue damage for the several selected details based on the process described along the following sections are carried out using in-house program FATA. BV uses generally its own SN curve with coefficient to account angle and welding direction, which is very similar to curve E. However typical SN curves like curves C, D and E may be used if acceptable for the hot spot location.

12 Wave scatter diagrams The trading history of the vessel is analyzed and organized in about ten major routes around the globe. However the history of some ships with regular voyages, as shuttle tanker for example, may be correctly represented with a few routes. Based on review of the tanker trading history, the representative routes, including transit voyage from the conversion yard to the final site location, are incorporated within the tanker phase analysis and represented with several zones from which the scatter diagrams are extracted from the BMT Global Wave Statistic database. The scatter diagram for the site has to represent several years of data. The JONSWAP formulation is generally adopted as the wave spectrum to perform the tanker phase analysis. Regarding the FPSO phase the site data and wave spectrum has to be respected. The Ochi- Hubble spectrum for example is very often used for Brazil or West Africa. Figure 9: Global wave statistic zones

13 Applied corrections Rainflow correction: In the fatigue methodology the short term variation of stress is assumed to be a narrow banded Gaussian process. However the actual variation of stress at a particular location might be a broad banded process. Because it is assumed that a fatigue damage of broad banded process is overestimated by the damage from an equivalent narrow banded process, this conservatism can be corrected. As a consequence some rainflow correction may be applied. Wave spreading function Directional spreading correction is generally applied in accordance with the IACS recommendation No.34. In this case the following spreading function is IACS recommendation No 34:

14 Intermittent wetting Ship motion analysis based on linear theory will not predict the non linear effect near the waterline due to intermittent wetting. In other terms the pressure accounted for details such as longitudinals on the side shell in the FEM calculation around the waterline is higher than the real one. In order to reduce the conservatism the stress due to this overestimated pressure is to be removed from the calculated RAOs. For this the global mesh is refined in the considered area and the fine mesh is rerun with the pressure described from the non linear theory. Different values are used for some parameters (such as the wave height) in order to describe the behavior of the correction on the different loading conditions and different waves applied on the ship. Figure 10: Example of intermittent wetting correction

15 Typical analyzed areas: The following areas are typically analyzed: - Longitudinal connections in way of adjacent frame to OT bulkhead : Bottom longitudinals, side shell longitudinals, deck longitudinals - Longitudinal connection in way of OT bulkheads: Bottom longitudinals, side shell longitudinals, deck longitudinals - End of horizontal girders - Typical connections on the transverse web frame - Topside supports along the vessel Figure 11: Examples of typical refined areas

16 Figure 12: Examples of typical refined areas

17 Conclusion: The deterministic approach was used to evaluate the fatigue life of trading tankers and then of offshore unities. This method was generally over conservative. The accuracy of the spectral fatigue is first due to the number of headings and frequencies accounted in the calculation. The frequencies sequence is refined in the resonance area. This allows us to catch the critical load cases which are not analyzed with the deterministic method. In addition the length of the wave along the vessel creates differences of pressures on the side shell. The real pressure along the ship creates real combinations of loads, such as pressures, vertical bending moments, horizontal moments and global shear force. It has to be noticed that the vertical bending is generally correctly accounted in the deterministic fatigue, but the shear force is not targeted in the 3 holds model and then the shear force is left free in the simplified method. As a similar way the correct associated horizontal bending moment and torsion effect are here perfectly applied for each load case. With the spectral fatigue it is possible to model the correct combination of swell and wind sea for bimodal configuration. The BV ARIANE software is able to calculate the orientation of the FPSO (for turret FPSO) for each sea state and then allow us to apply the correct heading for each wave component. Finally the BV FATA software is able to provide the final damage, accounting several corrections such as wave spreading function, rainflow methodologies and using any SN curves which may be required. Due to the number of load cases and to the required panel size (typically one element between stiffeners) the size of the databases could reach several gigaoctets. In order to improve the manipulation of the databases some innovative methodologies are under development. A procedure using a full ship model for the hydrodynamic calculation but only a three holds model for the structural model is under testing. Concentrated masses for the fore and aft part of the vessel are linked to the three holds model using rigid elements. Pressure integration gives nodal forces on FE model hull and concentrated forces and moments on the concentrated masses.

18 Pressure integration Figure 13: Methodology for 3 holds spectral fatigue

19 Global mesh model Hydrodynamic mesh VeriSTAR Structural mesh HydroSTAR VeriSTAR Dynamic Pressures Stillwater load case Interface HydroSTAR VeriSTAR Thousands of Load cases VeriSTAR: Refined mesh calculation Δσ For each Loading condition, Heading and Frequency RAO FATA5 Scatter diagram site meteocean data DAMAGE Figure 14: summary of the methodology

20 References: [1] BUREAU VERITAS Rules and regulation for the classification of ships, April 2007, [2] Offshore Rules, NI 493, 2007 [3] IACS recommendation No. 34 Standard Wave Data, [4] NI 393 July 1998, Fatigue Strength Of Welded Ship Structures [5] Veristar CSM, Cf. BIGOT/DEV VeriStar CSM 1.3c le 16/03/2006 [6] User Guide Hydrostar Manual, February 2007 [7] FATA v5. User Guide Manual, November 2007