8/9/2019 Fatigue Analyses on a Jacket http://slidepdf.com/reader/full/fatigue-analyses-on-a-jacket 1/21 | Jacket Fatigue, Earthquake, Transportation Analyses in Sesam | Date: 29 October 2014| | Author: A.B.Berdal | www.dnvgl.com/software Page 1 JACKET FATIGUE, EARTHQUAKE AND TRANSPORTATION ANALYSES IN SESAM This document explains necessary steps when proceeding from static analysis of a jacket using Sesam to: Dynamic spectral (stochastic) fatigue analysis Earthquake analysis Transportation analysis The 4 legged jacket shown in the figure above is used as a reference example. To run through this example first do as explained in the document Walkthrough example Jacket_4Leg_Comprehensive.pdf . The following pages describe the workflow step by step with emphasis on steps demanding special attention.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
The procedure described herein is a guide in how to do some important steps in connection with the
above-mentioned types of analysis. This guide should, however, not be taken as a complete descriptionof all necessary steps for these types of analysis.
A workflow analysis process for the jacket is controlled by Sesam Manager and is illustrated below (read
columns from left to right).
2
STATIC ANALYSIS OF FIXED MODEL
A linear static analysis is performed by the activity GeniE_static_fixed_model . A model fixed at the
bottom of the legs is created, a wave load analysis in Wajac is run and a linear static analysis in Sestra isrun. Wajac and Sestra are run under the control of GeniE. The purpose of the analysis is to serve as a
basis of comparison for a subsequent non-linear structure-pile-soil interaction analysis.
3 FREE VIBRATION ANALYSIS OF FIXED MODEL
A free vibration (eigenvalue) analysis is performed by the activity GeniE_freevib_fixed_model . A
model fixed at bottom of the legs is created, added mass is computed by Wajac and a linear free
vibration analysis in Sestra is run. Wajac and Sestra are run under the control of GeniE. The purpose of
the analysis is to serve as a basis of comparison for a subsequent free vibration analysis of a model with
A non-linear structure-pile-soil analysis is performed by the activity GeniE_static_piled_model . A
model with piles and soil is created, a wave load analysis in Wajac is run and a non-linear structure-pile-soil interaction analysis in Sestra and Splice is run:
Gensod computes soil curves/stiffnesses,
Sestra reduces the jacket, i.e. eliminates all nodes not connected to the piles,
Splice solves the pile-soil interaction non-linearly,
Sestra retracks the jacket, i.e. computes displacements and forces in nodes not connected to the
piles.
Wajac, Gensod, Sestra and Splice are run under the control of GeniE. This is an ultimate limit state (ULS)
analysis.
5
CHANGE THE MODEL USED FOR STATIC ANALYSIS
Before proceeding with the analysis some considerations must be made. These considerations are
explained in the following.
A spectral fatigue analysis in Framework may be based on a quasi-static analysis (neglecting inertia
effects in the structure) or a dynamic analysis in Sestra. In either case the wave loads are harmonic and
represented by complex loads (real and imaginary parts) hence the structural analysis as well as the
wave load analysis are in frequency domain.
5.1
Quasi-Static Analysis
In the case of quasi-static analysis the model used for static (e.g. ULS) analysis can to a large extent be
used also for spectral fatigue analysis. There are a couple of provisions though:
There should be no loads defined in GeniE or whichever preprocessor is used to create the T#.FEM
files. Such static loads are irrelevant for fatigue analysis in Framework and will also cause a failure in
Framework if included. So either delete any such loads in GeniE or create a new version of the model
without loads.
The pile-soil model must be replaced by linear spring elements at the seabed since a non-linear pile-
soil analysis in Splice is incompatible with a linear frequency domain analysis. See 6 for how to find
appropriate linear spring stiffnesses.
5.2
Dynamic Analysis
In the case of dynamic analysis the model normally needs some additional changes. Typical changes are:
In a dynamic analysis the model must be dynamically sound meaning that the stiffness must
correspond with the mass on a detailed level (each single d.o.f.) as well as overall. In a static
analysis soft parts (e.g. due to simplified modelling or low quality mesh) may be accepted since too
high static displacements in local areas may simply be neglected. In a dynamic analysis, however,
incorrect representation of mass and stiffness for a detail will influence the overall results and render
the analysis worthless.
Explicit loads (point-, line- and surface loads) representing equipment and other dead weights in a
static analysis must be replaced by mass in a dynamic analysis. In GeniE proper modelling of
equipment for static analysis is by use of either the Equipment feature or the Weight list feature
rather than explicit loads. For a dynamic analysis these equipments and weight list items should be
represented as mass rather than loads in the FE model. In the Property dialog for the loadcase(s)
containing equipments or weight lists select ‘Represent Equipment as loadcase-independent mass’.
Use either the ‘Footprint-Mass’ , ‘Beams-And-Mass’ or ‘Vertical-Beams-And-Mass’ representation. The
‘Eccentric-Mass’ option is less suitable for structural dynamic analysis and should be avoided.
There should be no loads defined in GeniE. Note that a load including only equipments and/or weight lists does not cause any problem, ref. the
item above. When the ‘Represent Equipment as loadcase-independent mass’ option has been
selected the ‘loads’ are converted to masses and the loadcase essentially ceases to exist.
The pile-soil model must be replaced by linear spring elements at the seabed since a non-linear pile-
soil analysis in Splice is incompatible with a linear frequency domain analysis. See section 6 for how
to find appropriate linear spring stiffnesses.
For a dynamic analysis a reduction technique may be required if the model is big. Of the two
reduction techniques in Sestra, the Master-Slave and the Component Mode Synthesis, the former is
the more appropriate one for fatigue analysis. If the model is split into superelements (as may be
the case if the model has been used for static analysis) then a reduction technique will necessarily be
used. Whether the model is a single or multi-superelement model the 1st level superelements must
be modified by introducing master nodes (supernodes) spread out over the model. This is required toproperly represent the dynamic energy of the model. Master (super) nodes should be defined with
the following in mind:
o Define only the three translations as super and let the three rotations be free as the
contribution of the rotations to the dynamic energy is normally modest.
o In case of a multi-superelement model remember that all six degrees of freedom of the
supernodes defined for coupling superelements together should be super to fully couple them
(only the translations as super would result in hinge coupling).
o Distribute supernodes all over the model.
o Have more supernodes where the dynamic energy is expected to be high due to large
displacements and/or due to high mass.
o Select nodes with high stiffness to be super. (The Master-Slave technique involves lumping
of mass to the master (super) nodes so low stiffness would result in too large displacementsof these nodes.)
6
DETERMINE LINEAR SPRING STIFFNESS – SPECIAL SPLICE
ANALYSIS
A procedure for determining a linear spring stiffness idealisation of the non-linear pile-soil is described in
section 14, p 14.0.1, of the document ‘Splice Engineering Documentation’ ( the document file name is
Splice_ED.pdf and is part of the Splice user documentation). The procedure described in this document is
exemplified in Case E.501, p E.07, of the document ‘Splice Verification Report’ (also found as part of the
Splice user documentation). This procedure is followed below.
A typical wave must somehow be selected for this special Splice analysis aimed at finding an equivalent
linear stiffness matrix. The wave to select may be the one contributing the most to fatigue damage or,
alternatively, the one for which 50% of the damage occurs for smaller waves and 50% for bigger waves.
One can imagine some advanced procedure for selecting this wave but the easiest may be to simply
make a qualified guess and then verify this guess based on fatigue results from Framework (using for
example the DEFINE FATIGUE-DUMP command).
For the example jacket a wave of 6 m and 8 sec. with direction from north 270º (the jacket is more
vulnerable in this direction) is chosen. (No attempt at verifying this choice is made in this example.)
First, do an analysis (GeniE including Wajac, Sestra and Splice) in which the non-linearity of the pile-soil
interaction is accounted for. This is performed by the activity GeniE_piled_model_comparison in the
A free vibration analysis with the spring-to-ground stiffness found above is performed (activity
GeniE_freevib_springed_model ) to find the resonance frequencies of the jacket. Compared with themodel used for static analysis the model must be modified as explained in a previous section (Change
the model used for static analysis).
Added mass must be computed by running Wajac (with MASS command).
Run an eigenvalue analysis in Sestra using the Implicitly Restarted Lanczos (Multifront Lanczos) method.
Both Wajac and Sestra are run from within GeniE.
The first three eigenperiods found are 3.04356, 2.47933 and 1.27722. The wave periods to specify to
Wajac in the subsequent forced response analysis in frequency domain should include these eigenperiods
of the structure.
Mode shapes should be studied in Xtract (using animation if necessary) to reveal irrelevant (incorrect?)
modes. You can start Xtract from within GeniE.
8
FORCED RESPONSE ANALYSIS – FIND DAFS AND DO
SPECTRAL FATIGUE ANALYSIS
The model used for the dynamic forced response analysis (in frequency domain) is the same as the one
used for the previous free vibration analysis. A separate activity named
GeniE_fatigue_springed_model is still defined for this GeniE execution since this activity must export
the T2.FEM file. This GeniE activity is together with the other activities of the stochastic fatigue task
organised under the sequence named Stochastic_fatigue. After running the activity
GeniE_fatigue_springed_model ensure that the file T2.FEM is found in the repository.
Run Wajac for computing wave transfer functions and added mass. The execution is found in the activity
named Wajac_frequency_domain .
In the Wajac input set OPT3 = 2. on the OPTI command in order to create G1.SIF containing transfer
functions for base shear and overturning moment. Select wave periods (note that periods and not
frequencies are given as input on the FRQ commands) based on:
Eigenperiods of the structure
Cancellation and amplification effects (wave length equal half of leg distance, equal leg distance,
double of leg distance, etc.), such cancellation and amplification effects are, however, neglected in
this example
Where wave energy is high
Distribute periods more or less evenly in area of high wave energy
Ensure that the files L2.FEM , S2.FEM and G1.SIF are produced by the Wajac run and copied (by Sesam
Manager) to the repository.
Present base shear and overturning moment transfer functions by reading the G-file from Wajac (G1.SIF )
into Postresp. This is activity Postresp_glob_transf_func. Use the graph to verify (and possibly
correct) choice of wave periods. Note that you may print the transfer functions to a file and import thisinto Excel to get access to more graphing options. The Postresp commands for this are:
‘Set > Print’ and set Destination to File in the ‘Print Options’ dialog
‘Print > Response variable’ and select for instance transfer functions FORCE1 (base shear in X) and
FORCE2 (base shear in Y) in directions 0 deg. and 90 deg.
The default name of the print file from Postresp is in this case Postresp_glob_transf_func.LIS.
Now run the Sestra forced response dynamic analysis using the direct frequency response method
(activity Sestra_dynamic_frequency_domain ). Proportional damping (Rayleigh damping) has been
selected with proportional factors 0.1 for both the mass matrix and the stiffness matrix. These may be
unrealistically high and it is the user’s responsibility to give proper values for his case. Ensure that the
results file R2.SIN is found in the repository.
Now we want to compare the dynamic response with the static response. For this we can compare the
base shear or overturning moment. The static response is found in the G1.SIF file previously produced
by Wajac, see above.
To find the dynamic response we must use Prepost to extract dynamic forces in the spring-to-groundelements and put these in a G-file (G2.SIF to avoid name conflict with the existing G1.SIF ). See the
figure below (produced by Xtract) showing element numbers. The Prepost input is given in section 8.2.
Note that changing the model in GeniE (e.g. adding or deleting beams) may change the element
numbers of the spring-to-ground elements thus necessitating modification of the Prepost input. The
Prepost execution is done by the activity Prepost_extract_transfer_func. This activity has Pre- and
PostExecuteScripts that take care of the ‘non-default’ copying of R2.SIN to the Prepost workspace and
G2.SIF back to the repository. Ensure that G2.SIF is found in the repository.
In Postresp (activity Postresp_dynamic_transf_func) combine the transfer functions (the transfer
functions or response variables are termed GRES1, GRES2, etc. on the G2.SIF file produced by Prepost)
to overall base shear (add contributions from all legs) and overturning moment (add contributions from
all legs plus moment effect of vertical forces in legs). The Postresp input is given in section 8.3.
Note: To make this task simple make sure the coordinate system of the spring-to-ground elements is not
askew with the global coordinate system.
Comparison of transfer functions from Wajac and Sestra gives dynamic amplification factors (DAF) at
base level. See the graphs below comparing static and dynamic base shears in X and Y, respectively.
(The spreadsheet ‘Compare static and dynamic response.xlsx ’ producing the graphs is found together
with the input files.) Based on the graphs of the transfer functions you may want to revise the choice of
wave frequencies.
To get DAFs at various levels and not only at the bottom one must do a quasi-static analysis in Sestra
and compare this with the dynamic. This is briefly explained in section 8.1.
The model used in this case should be only the jacket without conductors and topside.
In a transportation analysis the jacket is resting on its side. It must therefore be rotated in GeniE. A way
to find the rotation angle (which will be more than 90º due to the sloping jacket panel) is as follows.
Create two auxiliary beams as shown in the figure below. The angle between them will be the rotation
angle for the jacket to make it rest on its side. (In GeniE angles between intersecting beams may be
labelled.) Deleted the two auxiliary beams are after the rotation.
The jacket is resting on its side on some supports on top of a barge. The sea-fastening is a set of
inclined beams supporting the jacket sideways and is welded to the jacket only after the jacket’s
deflection due to gravity. I.e. the sea-fastening is unloaded by the gravity of the jacket. The rolling of
the barge then subjects the jacket to rotational accelerations which put load on the sea-fastening.
Just for comparison purposes a simplified transportation analysis is performed by the activityGeniE_simple_transportation. In this analysis the effect of adding the sea-fastening after the jacket’s
deflection due to gravity is neglected. I.e. it is assumed that the sea-fastening is loaded by the deflection
due to gravity. Sestra is run under the control of GeniE.
A proper transportation analysis is then performed by the sequence named
Transportation_with_prestress.
The loading of the sea-fastening only after the jacket’s deflection due to gravity is achieved by the
following procedure:
First a static analysis with gravity load only is performed and with a very soft material assigned to
the sea-fastening. This is the two activities GeniE_transport_pre_seafast (producing the file
PreSeafastT4.FEM ) and Sestra_static_pre_seafast (producing the file PreSeafastR4.SIN ).
| Jacket Fatigue, Earthquake, Transportation Analyses in Sesam | Date: 29 October 2014 |
Then a new static analysis with rotational loads is performed, this time with normal material
properties for the sea-fastening. This is the two activities GeniE_transport_seafast (producing the
file SeafastT4.FEM ) and Sestra_static_seafast (producing the file SeafastR4.SIU ).
These two result files are then merged using Prepost. This is activity Prepost_merge_transport. APreExecuteScript named CopyResultsFileTransport.js first copies PreSeafastR4.SIN into MergeR4.SIN
before SeafastR4.SIU is merged into it.
In Xtract the gravity result case is combined with the rotational result cases. This is activity
Xtract_present_results. The figure below shows deflections and axial forces in the beams for this