Professional Documents
Culture Documents
Lorenzo Marchionni and Luigino Vitali Saipem Energy Services S.p.A. Fano (PU), Italy Email: lorenzo.marchionni@saipem.eni.it and luigino.vitali@saipem.eni.it
Summary
Offshore pipelines are generally subject to lateral and upheaval buckling instability under operating conditions due to internal pressure loads and thermal expansion. In addition, deep water offshore pipelines can be subject to lateral buckling during system pressure test caused by the high compressive loads applied on the steel during installation and high internal pressure applied during the system pressure test. The lateral buckling is a complex problem with a lot of related parameters and is very time expensive to perform a reverse engineering activity well knowing the history load but without to have exact informations about soil properties affecting pipe-soil non-linear resistance forces one of the main parameters governs the actual behaviour of offshore pipelines under lateral buckling. This paper refers to the experience on an offshore pipeline laid in deep water and experiencing lateral buckling during the system pressure test. The multi-objective optimisation was used to obtain the best value combination of the un-known relevant parameters. ModeFrontier (v. 4.0.2) has been used to perform the optimisation process. In this paper, the following is discussed: Brief description of the phenomenon; Description of the FE model to simulate the pipeline response during system pressure test and operating conditions; Selection of the un-known relevant parameters and of design targets: relevant parameters are related to the pipe-soil resistance force, while design targets are related to the pipe configuration and maximum deformation; The results of the FE model parametric study carried out to fit the measured lateral pipeline configuration using the classical trial and error optimisation technique; Description of the optimisation processes used in ModeFrontier and selection of the Pareto curve; Comparison of the output values subsequently the optimisation process with respect to the design targets and considerations about the saving time calculation and the results accuracy achieved.
Keywords
Pipeline, Lateral Buckling, Multi-Objective Optimisation.
Introduction
When a pipeline resting on the seabed is operated at high pressure and high temperature, it develops lateral bending to recover thermal expansion. If the line is restrained by soil friction, axial compressive load develops for increasing temperatures and, at some critical value, the pipe may experience lateral deformation/bending paths activated by initial misalignments. As soon as the pipeline starts deflecting laterally, the initial resistance of the line to develop lateral deflection is reduced at such locations. Depending on how the thermal expansion/restraint develops, the pipeline may slip from adjacent sections into one isolated buckle. An equilibrium is achieved when the mobilised friction equilibrates the axial force in the buckle. The axial compressive force within the buckle depends on the length and modal shape. At equilibrium achieved, bending at worst localisation/crest of the buckle might be excessive. The release of the axial compressive force as a consequence of the lateral bending is gradual from an anchoring point through the two adjacent lengths feeding into the buckle. In presence of low axial friction, the two lengths adjacent to the buckle are very long, so they give rise to a large feed-in potentially causing unacceptable bending. Any perturbation from the rectilinear configuration may cause the pipeline to buckle laterally and to achieve a new shape, but on the horizontal plane. In the early years, a large number of deep water offshore pipelines have been laid. During system pressure test, the pipeline is flooded with sea water and an internal pressure 10%-15% larger than the design pressure is applied. During this phase the pipeline can be subject to lateral buckling due to the high compressive loads applied on the steel during installation and high internal pressure. Pipeline technology started to account for is-service buckling in the early eighties. A series of papers, e.g. Hobbs (1984, Ref. [5]) and Taylor and Ben Gan (1986, Ref. [6]), proposed analytical tools to predict the occurrence and the consequence of in-service buckling, see Figure 1. Pipe buckle formation is governed by three parameters: Pressure and temperature gradient giving rise to the effective axial compression in the pipeline; Pipeline initial configuration in the horizontal plane affecting the buckling mode shape, see Figure 2; Non-linear pipe-soil interaction relationship in the axial and lateral direction.
Although only three parameters are involved, a number of factors feed into each parameter and there is significant uncertainty over the true magnitude of each. A buckle forms when the local combination of driving force and OOS exceed a critical value. The response of a pipeline system to various imperfection sizes is illustrated in Figure 1 (Ref. [6]).
Figure 1
Figure 2
The FE structural modelling is rather difficult and involves soil-structure interaction mechanisms, unknown frictional loads, pipe plasticization effects, large deflections, etc. The uncertainty in the definition of parameters that characterises the soil-pipe interaction, together with a rough knowledge of the actual equilibrium configuration that the pipe assumes on the seafloor often require the use of simplified interpretative/predictive models combined with sensitivity analysis rather than sophisticated models. This paper describes the experience on an offshore pipeline laid in deep water and experiencing lateral buckling during the system pressure test. A FE model has been performed to evaluate the pipeline response under system pressure test conditions and the pipe-soil interaction forces have been calibrated to obtain both the measured pipeline configuration in the horizontal and vertical plane and the maximum post-buckling deformation. The multi-objective optimisation has been carried out using ModeFrontier (v. 4.0.2) to obtain the best value combination of the un-known relevant parameters. The paper concludes with the comparison of results obtained using the traditional trial and error technique and the optimisation software.
Nomenclature
FEM FEA SPT DOE MOGA Finite Element Model Finite Element Analysis System Pressure Test Design of Experiment Multi Objective Genetic Algorithm
In the following the development of the axial compression force is detailed. The first loading condition is the one residual from the laying phase (Figure 3a) and the axial force applied on the steel pipe wall is:
NS,Lay = NR,Lay p e A e
(1)
where NR,Lay is the residual axial force on the sea bottom from laying, pe is the hydrostatic external pressure and Ae is the external steel pipe cross section area.
Being the system pressure test performed with water, the axial force applied on the steel pipe wall after pipe flooding changes (Figure 3b). Assuming a fully axially restrained condition, the steel pipe wall after pipe flooding increases, Ns,flood, as follows:
NS,Flood = + H (p e ) A S = +
pe D AS 2t
(2)
where is the Poisson coefficient, D is the pipe steel diameter (average diameter Dm can be used), t is the pipe steel thickness, H is the applied hoop stress variation from the as-laid condition. When applying the pressure for testing the system, assuming a fully axially restrained condition, the steel pipe wall after pipe reaching the maximum test pressure (Ptest at reference height) changes, NS,test, as follow (Figure 3c):
NS,Test = + H (p test ) A S = +
p test D AS 2t
(3)
where H is the applied hoop stress variation from the flooded pipe condition. According to the above the total axial force on the steel pipe wall during the system pressure test is:
= NR,Lay p e A e +
pe D p D A S + Test AS 2t 2t
(4)
The effective steel axial force, NE,Test, is then calculated according to equation (5.3) in DNV OS-F101 clause 209 section 5, and reads:
(5)
D A S (p e + p Test ) A i 2t
(6)
Assuming AiAeAm (Am is the mean pipe cross section area) and calculating the steel cross section area AS as Dmt (Dm being the mean diameter), equation (5.4) in DNV OS-F101 clause 209 section 5 is readily obtained:
(7)
In the above, it assumed that the temperature change with respect to the as-laid condition during system pressure test is negligible. From equations (6) and (7), it is evident that, for very deep water pipelines, the effective compressive axial force is larger than for shallow water application of the same pipe. In fact, the difference in compression is related to the term pe: the larger the water depth the larger the effective axial compression.
NR,Lay NS,Lay
pe
NS,Floo
d
(b) pe
pi=pe
NS,Test
NS,Test
Figure 3
Scheme of the applied external loads and axial forces on the pipeline during the different construction phases: (a) installation, (b) flooded, (c) and (d) system pressure test.
Out-of-straightness
GEOPIG Survey Data vs. FEM Analysis 5.0
4.0
3.0
Out-of-straight (m)
2.0
Lateral Configuration
1.0
0.0
-1.0
-2.0
-3.0 84500
84525
84550
84575
84600
84625
84650
84675
84700
84725
84750
84775
84800
84825
84850
84875
84900
84925
84950
84975
85000
85025
85050
85075
85100
KP (m)
-2128.50
-2128.75
-2129.00
-2129.25
-2129.50
-2129.75
-2130.00
-2130.25 84500
84525
84550
84575
84600
84625
84650
84675
84700
84725
84750
84775
84800
84825
84850
84875
84900
84925
84950
84975
85000
85025
85050
85075
85100
KP (m)
Figure 4a
0.4
0.3
Longitudinal Strain
0.2
0.1
-0.1
-0.2 84650 84675 84700 84725 84750 84775 84800 84825 84850 84875 84900 84925 84950
X-coord (m)
Figure 4b
FE Model
Finite Element Analyses using ABAQUS Code (Rel. 6.7.1) have been carried out to assess pipe response of the 24 pipeline, under the effect of pressure test loads and laid on an irregular seabed. The FE model accounts the development of a single buckle starting from a straight pipeline section. The model length is sufficient to anchoring the pipeline on both pipeline ends. The following aspects have been included in the FE model (see also Refs. [1] and [2]): Steel material non linearit; Large rotation; Non-linear soil-pipe interaction; Flat sea bottom, modeled using R3D4, 4 nodes plane element, assumed as uniform transversally to the pipeline axis.
The FE analyses carried out have been performed modeling the pipeline using PIPE elements (ABAQUS PIPE31 Element). This element is able to predict global pipeline behavior. All the PIPE elements are two diameters long. Coulomb friction contact elements have been used to simulate both lateral and axial pipe soil interaction forces. The seabed profile has been considered as flat. Pressure load is specified as element load. The internal pressure effects on the pipe wall (Poisson effect) and 3-D stress material formulation (hoop and longitudinal stress) are accounted in the analysis.
Loading Sequence
The FE model is subject to different load steps to simulate the different load condition the pipeline will be subject to during the SPT, particularly: Step 1: Apply pipe submerged weight and lay the pipeline section on the irregular sea bottom (a smoothed vertical configuration has been used to model the seabed) considering the pipe rectilinear in the un-deformed configuration; Step 2: Apply external pressure; Step 3: Generate lateral imperfection; Step 4: Release boundary conditions; Step 5: Apply pipe submerged weight under pressure test; Step 6: Apply internal pressure test; Step 7: Release hydrotest load conditions; Step 8: Apply pipe submerged weight under operating conditions; Step 9: Apply internal operating pressure.
Effective axial force; Minimum (compressive) and maximum (tensile) value of longitudinal/axial strain.
For the different FEM cases analyzed, the distribution of lateral displacements, steel axial forces, bending moments, maximum tensile strains and minimum compressive strains in proximity of the purpose-made features does not change significantly as the pressure loads and thermal expansion increases. Exception is made for the extension of the imperfection length interested by lateral snaking and magnitude of maximum lateral displacement, maximum steel axial force, maximum tensile strain and minimum compressive strain.
Figure 5
Lateral Reaction
4 3.5
3 2.5 2 1.5 1 0.5 0 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3
Outside of Buckle Inside of Buckle Shoulder of Buckle
Axial Reaction
0.35 0.3
0.25 0.2 0.15 0.1 0.05 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Inside of Buckle
Figure 6a
8 Downward Reaction (kN) 7 6 5 4 3 2 1 0 0
Outside of Buckle
Inside of Buckle
Shoulder of Buckle
0.2
0.4
0.6
1.4
1.6
1.8
Uplift Reaction
2.5
1.5
0.5
Outside of Buckle
Inside of Buckle
0 0 0.1 0.2 0.3 0.4 0.5 0.6 Vertical displacement (m) 0.7 0.8 0.9 1
Figure 6b
Table 1 lists the input variables used in the sensitivity analysis and specify the range of variation and the minimum step considered in the first optimisation process. Input Data Axial Soil Friction factor Lateral Soil Friction factor Maximum Relative Displacement for the Full Friction Mobilization Initial Lateral Trigger Table 1 Input Variables. Range of Variation 0.14 0.18 0.7 1.0 0.5m 1.0m 0.3m 1.0m Step of Variation 0.0025 0.005-0.01 0.025m 0.025m
The pipeline response has been evaluated monitoring the maximum lateral displacement (see Figure 7) and the maximum longitudinal bending strain (see Figure 8) at the apex of the secondary and main wave (areas #3 and #4). In particular, the following design targets have been fixed:
Output Data Max. Lateral Displacement in the area #3 Max. Lateral Displacement in the area #4 Max. Longitudinal Strain in the area #3 Max. Longitudinal Strain in the area #4 Table 2 Design Targets.
FEM Results
A sensitivity FEM analysis has been carried on the main relevant parameters to identify the best combination of values to obtain the fitting of the measured pipeline configuration and the maximum longitudinal deformation.
Out-of-straightness
GEOPIG Survey Data vs. FEM Analysis 5.0
Out-of-straight (m)
3.0
0.0
-1.0
-2.0
-3.0 84500
84525
84550
84575
84600
84625
84650
84675
84700
84725
84750
84775
84800
84825
84850
84875
84900
84925
84950
84975
85000
85025
85050
85075
85100
KP (m)
Figure 7
0.5 0.4
Pipeline Configuration.
Longitudinal Strain GEOPIG Survey Data vs. FEM Analysis
FEM Analysis - Operating Conditions FEM Analysis - Design Conditions Survey Data - Smoothed 1
0.3 0.2
X-coord (m)
Figure 8
The variation of the un-known relevant parameters listed in Table 3 was controlled using the traditional trial and error technique. The FEM analysis results are summarized as follows:
Output Max. Lateral Displacement in the area #3 Max. Lateral Displacement in the area #4 Max. Longitudinal Strain in the area #3 Max. Longitudinal Strain in the area #4 Table 3 Design Targets vs. FEM Results.
10
Multi-Objective Optimisation
The multi-objective optimisation has been carried out using the commercial software ModeFrontier (v.4.0.2). Two optimisation processes have been performed to obtain sufficient results accuracy. In the second run, the range and step of variation of any input data have been reduced aiming to obtain better results accuracy. The Design of Experiment (DOE) algorithm used to generate the initial design group was the SOBOL (see Figure 9). This technique permitted to obtain better distributed designs (suitable for a low number of input variables, <10).
Figure 9
At the end of the base designs run, the optimisation algorithm used to find the solutions lie in the Pareto front was the MOGA-II (Multi Objective Genetic Algorithm). MOGA-II is an improved version of MOGA by Poloni and its efficiency is ruled by its operators (classical crossover, directional crossover, mutation and selection) and by the use of elitism (elitism ensures that the best solutions are preserved during the evolutions), see Figure 10.
Figure 10
The following techniques have been used in the optimisation runs: 1 Optimisation process: Design DOE: 32 designs SOBOL DOE; Optimisation algorithm: 18 generations MOGA-II; Total number of FE designs: 576;
11
Time for any FEM simulation: 10-12min; Global time of calculation: 96h. Design DOE: 35 designs SOBOL DOE; Optimisation algorithm: 19 generations MOGA-II; Total number of FE designs: 683; Time for any FEM simulation: 10-12min; Global time of calculation: 120h.
2 Optimisation process:
Results Comparison
The classical sensitivity FEM analysis and the optimisation processes using ModeFrontier gave sufficient results accuracy performing a good value combination of input data. No further optimisation processes are necessary.
PARETO CURVE
Figure 11
The optimisation analysis results using ModeFrontier are summarized as follows: Output Max. Lateral Displacement in the area #3 Max. Lateral Displacement in the area #4 Max. Longitudinal Strain in the area #3 Max. Longitudinal Strain in the area #4 Table 4 Design Targets vs. Optimisation Results. Optimisation Results 1.76m 4.44m 0.156% 0.403% Target 2.08m 4.50m 0.156% 0.40%
12
Conclusions
The reverse engineering analysis has been carried out before using the classical trial and error technique and then using the commercial software ModeFrontier (v. 4.0.2). In the latter case, two optimisation processes have been performed starting from a large range of variation on the input data. Comparing the results, the following conclusions can be deduced: The results obtained at the end of the two optimisation techniques show similar input data values and the difference between the design targets and output data physically comparables (see Table 3 and Table 4); Using the trial and error technique 13-15 workdays (105-120 manhours) have been necessary. While using ModeFrontier, the two optimisation processes have been concluded in 4-5 workdays (30-40 manhours).
The main advantage in the use of ModeFrontier has been the remarkable reduction in the calculation time that it is comparable at the 50% if we consider only the workdays and at the 100% if we consider the holidays. This time reduction had not an influence in the results accuracy.
References
[1] Vitali L., Spinazz M., Verley R. (1999): The HOTPIPE Project Use of Analytical Models/Formulas in Prediction of Lateral Buckling of Isolated and Interacting Buckles, Proc. 9th Int. Offshore and Polar Engineering Conference, 1999; Collberg L. et al. (2005): HotPipe JI Project - Design Guideline for High Temperature/High Pressure Pipelines, OMAE Paper No. 67523, Proc. 24th OMAE Conference, Halkidiki, Greece, 12-17 June 2005. Spinazz M. Torselletti E. & Levold E.(1999): The HOTPIPE Project A Study of the Effectiveness of Remedial Measures to tackle/control the Development of Excessive Bending, Proc. 9th Int. Offshore and Polar Engineering Conference, 1999. Torselletti E., Vitali L. & Levold E. (1999): The HOTPIPE Project Snaking of Submarine Pipelines Resting on Flat Sea Bottom using Finite Element Method, Proc. 9th Int. Offshore and Polar Engineering Conference, 1999. Hobbs, R. E.: Pipeline Buckling Caused by Axial Loads, Journal of Constructional Steel Research, Vol. 1, No. 2, January 1981. Taylor, N. and Ben Gan, A.: Submarine Pipeline Buckling Imperfection Studies, Thin-walled Structures, 4, 295-323, 1986. Spth H.: Spline Algorithms for Curves and Surfaces.
[2]
[3]
[4]
13