Documente Academic
Documente Profesional
Documente Cultură
DNV-RP-C208
Draft 3 November 2012
— CONFIDENTIAL
— This document is a draft not intended for external distribution
DRAFT
TABLE OF CONTENT
1 INTRODUCTION 5
1.1 PURPOSE OF THE DOCUMENT 5
1.2 VALIDITY 5
1.3 DEFINITIONS 5
1.4 SYMBOLS 6
2 BASIC CONSIDERATIONS 7
2.1 LIMIT STATE SAFETY FORMAT 7
2.2 CHARACTERISTIC RESISTANCE 9
2.3 TYPE OF FAILURE MODES 9
2.4 USE OF LINEAR AND NON-LINEAR ANALYSIS METHODS 9
2.5 EMPIRICAL BASIS FOR THE RESISTANCE 9
2.6 DUCTILITY 10
2.7 SERVICEABILITY LIMIT STATE 10
2.8 PERMANENT DEFORMATIONS 10
3 GENERAL REQUIREMENTS 10
3.1 DEFINITION OF FAILURE 10
3.2 MODELLING 11
3.3 DETERMINATION OF CHARACTERISTIC RESISTANCE TAKING INTO ACCOUNT STATISTICAL VARIATION
11
3.4 REQUIREMENT TO THE SOFTWARE 12
3.5 REQUIREMENT TO THE USER 12
4 REQUIREMENTS TO THE FEM ANALYSIS 12
4.1 GENERAL 12
4.2 SELECTION OF FE SOFTWARE 12
4.3 SELECTION OF ANALYSIS METHOD 12
4.3.1 Implicit versus explicit solver 12
4.3.2 Solution control for explicit analysis 13
4.3.3 Solution control for dynamic implicit analysis 13
4.3.4 Solution control for static implicit analysis 14
4.4 SELECTION OF ELEMENTS 14
4.5 MESH DENSITY 15
4.5.1 General 15
4.5.2 Mesh refinement study 15
4.6 GEOMETRY MODELLING 15
4.7 MATERIAL MODELLING 16
4.7.1 General 16
4.7.2 Material models for metallic materials 16
4.7.3 Stress strain measures 17
4.7.4 Evaluation of strain results 18
4.7.5 Stress - strain curves for buckling and ultimate capacity analyses 19
4.7.6 Strain rate effects 21
4.8 BOUNDARY CONDITIONS 22
4.9 LOAD APPLICATION 22
4.10 APPLICATION OF SAFETY FACTORS 22
4.11 EXECUTION OF NONLINEAR FE ANALYSES, QUALITY CONTROL 22
4.12 REQUIREMENTS TO DOCUMENTATION OF THE FE ANALYSIS 23
5 REPRESENTATION OF DIFFERENT FAILURE MODES 23
5.1 DESIGN AGAINST TENSILE FAILURE 23
5.1.1 General 23
5.1.2 Tensile failure resistance from non-linear analysis calibrated against a known solution 24
5.1.3 Tensile failure in base material. Simplified approach. 24
1 INTRODUCTION
1.1 Purpose of the document
This document is intended to give guidance on how to establish the structural resistance by use of non-
linear FE-methods. It deals with determining the characteristic resistance of a structure or part of a
structure in a way that fulfils the requirements in DNV offshore standards.
Non-linear effects that may be included in the analyses are e.g. material and geometrical non-linearity,
contact problems etc.
The characteristic resistance should represent a value that meet the requirement that there is less than
5% probability that the resistance is less than this value. This definition of the characteristic resistance
is similar to what is required by many other structural standards using the limit state safety format and
these recommendations are expected to be valid for determination of capacities to be used with such
standards.
The recommendations are foreseen to be used for cases where the characteristic resistance is not
directly covered by codes or standards. The goal is that analyses carried out according to the
recommendations given in this document will lead to a structure that meets the requirements to the
minimum safety margin in the standard.
This document is not intended to replace formulas for resistance in code and standards for the cases
where they are applicable and accurate, but to present methods that allows for using non-linear FE-
methods to determine resistance for cases that is not covered by codes and standards or where accurate
recommendations are lacking.
1.2 Validity
The document is valid for marine structures made from structural steels meeting requirements to
offshore structures with a yield strength of up to 500 MPa.
The recommendations presented herein are adapted to typical offshore steels that fulfil the
requirements specified in DNV-OS-C101 or an equivalent offshore design standard. The specified
requirements are made under the assumption that the considered structure is operating under
environmental conditions that are within the specification of the applied offshore standard. If the
offshore unit is operating outside these specifications, the failure criterion presented in this
recommended practice can only be utilized if it can be documented that both the weld and parent
material has sufficient toughness in the actual environmental conditions.
1.3 Definitions
This Recommended Practise use terms that is defined in DNV-OS-C101. The following additional
terms are defined below:
conservative load load that maintain is orientation when the structure deforms (e.g. gravity loads)
characteristic the resistance that for a particular failure mode is meeting the requirement of having a
resistance prescribed probability that the resistance falls below this value (usually 5% fractile)
ductility the ability to deform beyond the proportionality limit without significant reduction in the
capacity due to fracture or local buckling (originally ductility refer to the behaviour of
the material, but is here also used for the behaviour of structures and structural details)
follower load load that change direction with the structure (e.g. hydrostatic pressure)
gross yielding yielding across the entire structural component e.g. a flange
1.4 Symbols
b span of plate
c flange outstand
C damping matrix
D Outer diameter of tubular sections
E modulus of elasticity
𝐸𝑝1 stress-strain curve parameter
𝐸𝑝2 stress-strain curve parameter
𝐹𝑒𝑥𝑡 external forces
𝐹𝑖𝑛𝑡 internal forces
𝑓𝑦 yield stress/yield strength
K Ramberg-Osgood parameter
𝑘𝑔 eigenvalue for governing buckling mode
𝐿𝑠 characteristic element size of smallest element
M mass matrix
N number of cycles to failure
𝑅𝑑 design resistance
𝑅𝑘 characteristic resistance
𝑆𝑑 design action effect
𝑆𝑘 characteristic action effect
t time, thickness
u displacement vector
𝜀 strain
2 BASIC CONSIDERATIONS
2.1 Limit state safety format
A limit state can be defined as: “A state beyond which the structure no longer satisfies the design
performance requirements.” See e.g. /1/.
Ultimate limit states (ULS) corresponding to the ultimate resistance for carrying loads.
Fatigue limit states (FLS) related to the possibility of failure due to the effect of cyclic loading.
Accidental limit states (ALS) corresponding to failure due to an accidental event or
operational failure.
Serviceability limit states (SLS) corresponding to the criteria applicable to normal use or durability.
This document deals with limit states that can be grouped to ULS and ALS. This standard does also
address failure modes from cyclic loading for cases that cannot adequately be checked according to
the methods used in codes for check of FLS. This is relevant for situations where the structure is
loaded by a cyclic load at a high load level but only for a limited number of cycles (low-cycle fatigue).
The safety format that is used in limit state codes is schematically illustrated in Figure 2-1.
Sd< Rd
Sd ≤ Rd
(1)
The characteristic resistance should represent a value which will imply that there is less than 5%
probability that the resistance is less than this value.
The characteristic resistance given in design codes is determined also of considerations of other
aspects than the maximum load carrying resistance. Aspects like post-peak behaviour, sensitivity to
construction methods, statistical variation of governing parameters etc. are also taken into account. In
certain cases these considerations are also reflected in the choice of the material factor that will be
used to obtain the design resistance. It is necessary that all such factors are considered when the
resistance is determined by non-linear FE-methods.
2.3 Type of failure modes
When steel structures are loaded to its extreme limit they will either fail by some sort of instability
(e.g. buckling) that prevent further loading or by tension failure or a combination of the two. For
practical cases it is often necessary to define characteristic resistance at a lower limit in order to be
able to conclude about structural integrity without excessive analysis. Examples of this can be limiting
the plastic strain to avoid cyclic failure for dynamically loaded structures or deformation limit for
structural details failing by plastic strain in compression. See Section 3.1.
The following types of failure modes are dealt with in this Recommended Practise:
• Tensile failure
• Failure due to repeated yielding (low cycle fatigue)
• Accumulated plastic strain
• Buckling
• Repeated buckling
2.4 Use of linear and non-linear analysis methods
Traditionally, offshore structures are analysed by linear methods to determine the internal distribution
of forces and moments, and the resistances of the cross-sections are checked according to design
resistances found in design codes. These design resistance formulae often require deformations well
into the inelastic range in order to mobilise the code defined resistances. However, no further checks
are normally considered necessary as long as the internal forces and moments are determined by linear
methods. When non-linear analysis methods are used, additional checks of accumulated plastic
deflections and repeated yielding will generally be needed. These checks are important in case of
variable or cyclic loading e.g. wave loads.
2.5 Empirical basis for the resistance
The integrity of a structure is also influenced by other factors than the value of the characteristic
resistance. The ability of a structural detail to maintain its resistance in case of overload is highly
influencing the resulting reliability of the structure. It is therefore necessary to consider not only the
value of the resistance when determining the characteristic resistance, but also to judge how the load
deflection relationship is for a particular failure mode.
The check for ductility requires that all sections subjected to deformation into the inelastic range
should deform without loss of the assumed load-bearing resistance. Such loss of resistance can be due
to tensile failure, instability of cross-sectional parts or member buckling. The design codes give little
explicit guidance on this issue, with exception for stability of cross-sectional parts in yield hinges,
which normally are covered by requirement to cross-sectional class 1. See e.g. DNV OS-C101 /13/.
Steel structures behave generally ductile when loaded to their limits. The established design practise is
based on this behaviour, which is beneficial both with respect to simplifying the design process and
improving the performance of the structure. For a ductile structure, significant deflections may occur
before failure and thus give a collapse warning. Ductile structures also have larger energy absorption
capabilities against impact loads. The possibility for the structure to redistribute stresses lessens the
need for an accurate stress calculation during design as the structure may redistribute forces and
moments to be in accordance with the assumed static model. This is the basis for use of linear analyses
for ULS checks even for structures, which behave significantly non-linear when approaching their
ultimate limit states.
2.7 Serviceability Limit State
Use of non-linear analysis methods may result in more structural elements being governed by the
requirements to the Serviceability Limit State and additional SLS requirements may be needed
compared with design using linear methods. E.g. when plate elements are used beyond their critical
load out of plane deflections may need to be considered from a practical or aesthetic point of view.
2.8 Permanent deformations
All steel structures behave more or less non-linear when loaded to their ultimate limit. The formulas
for design resistance in DNV Offshore standard /13/ or similar codes and standards are therefore
developed on the basis that permanent deformation may take place before the characteristic resistance
is reached.
3 GENERAL REQUIREMENTS
3.1 Definition of failure
In all analyses a precise definition of failure should be formulated. The failure definition needs to
correspond with the functional requirement to the structures. In certain cases like buckling failure it
may be defined by the maximum load, while in other cases it need to be selected by limiting a suitable
control parameter e.g. plastic strain.
For Ultimate Limit States (ULS) and Accidental Limit States (ALS) the definition of failure needs to
reflect the functional requirement that the structure should not loose is load-carrying resistance during
the dimensioning event. That may e.g. imply that in an ULS check the failure is defined as the load
level where the remaining cycles in the storm that includes the ULS loadcase, will not lead to a
progressive or cyclic failure unless a specific check for these failure modes are carried out. See
also 5.2. Another example is in case of an ALS check for blast pressure one may consider the failure
criterion to be the limiting deflection for the passive fire protection.
It shall be checked that the analysis tool and the modelling adopted represent the non-linear behaviour
of all structural elements that may contribute to the failure mechanism with sufficient accuracy. The
model should be suitable to represent all failure modes that are intended to be checked by the analysis.
It should be made clear which failure modes the model will adequately represent and which failure
modes that are excluded from the analysis and that are checked by other checks.
3.3 Determination of characteristic resistance taking into account statistical variation
When FE-methods are used to determine the structural resistance it is necessary to take due account of
the statistical variation of the various parameters such that the results will be equal to or represent an
estimate to the safe side compared with what would be obtained if physical testing could be carried
out.
The model should aim to represent the resistance as the characteristic values according to the
governing code. In general that means 5 % fractile in case a low resistance is unfavourable and 95 %
fractile in case a high resistance is unfavourable.
In cases where data of the statistical variation of the resistance is uncertain one need to establish a
selection of the governing parameters by engineering judgement. The parameters should be selected
such that it can be defended that the characteristic resistance established meets the requirement that
there is less than 5% probability that the capacity is below this value.
All parameters that influence the variability of the resistance need to be considered when establishing
the characteristic resistance.
It is therefore necessary to validate the analysis procedure according to one of the following methods:
a) Selection of all governing parameters to be characteristic or conservative values.
In this method all parameters that influence the result (key parameters) are selected to give
results to the safe side. E.g. element type, mesh size, material curve, imperfections, residual
stresses etc.
For structures or structural details where the resistance is dominated by the value of the yield
stress, using the specified minimum yield stress according to offshore steel material standards
will represent the requirements to the characteristic resistance. Other parameters with
statistical variation that will influence the resistance e.g. plate thickness should be selected as
a safe estimate of the expected value in order to meet the required statistical requirement for
the resulting resistance. In cases of doubt a sensitivity assessment may be necessary.
In some cases values are given in the codes for analysis of specific problems see e.g. 5.4.3.
b) Validation against code values
In this method a selected code case is used for calibration (denoted code calibration case). The
case should represent the same failure mode that is to be investigated. The key parameters e.g.
element type, mesh size, material curve, imperfections, residual stresses etc. should be
selected so the analysis provide the resistance predicted by the code for the code calibration
case. The same parameters are then used when the resistance of the actual problem is
determined.
If the analysis is calibrated against ordinary code values that meet the requirements to
characteristic resistance then the resistance of the analysed structure also will meet the
requirement.
c) Validation against test
In this method one or more physical tests that is judged to fail in a similar way as the problem
to be analysed are selected for calibration (denoted test calibration case). First the key
parameters e.g. element type, mesh size, material curve, imperfections, residual stresses etc.
The software used shall be documented and tested for the purpose.
3.5 Requirement to the user
The user should be familiar with FE-methods in general and non-linear methods in particular.
The analysts need to understand the structural behaviour of the problem in question.
The user shall know the theory behind the methods applied as well as the features of the selected
software.
When documenting structures to meet a code described reliability level with use of non-linear methods
for determination of the resistance it is necessary the engineers understand the inherent safety
requirements of the governing code.
The term non-linear FE- analysis covers a large number of analysis types for different purposes and
objects. The content of this section is written with analyses of steel structures in mind. The objective is
to document structural capacity of the structure in a way that fulfils the requirements for determining
characteristic resistance in DNV offshore standards and other similar standards, such as the Norsok N-
series and the ISO 19900 suite of standards.
4.2 Selection of FE software
The software must be tested and documented suited for analysing the actual type of non-linear
behaviour. This includes:
• Non-linear material behaviour (yielding, plasticity)
• Non-linear geometry (Stress stiffening, 2’nd order load effects)
Both implicit and explicit equation solvers may be used to solve the general equation system:
Most explicit FE codes calculate the governing size of the time step based on equations similar to
Equation (3). For problems of longer duration, one often wants to save analysis time by reducing the
number of time steps. This can be done by accelerating the event or mass scaling. Accelerating the
event reduces the simulation time and thus computational time, the mass scaling increases the time
step reducing the computational time, see Equation (3).
The time saving methods are only useful if the inertia forces are small. Thus, it must be documented
that the kinetic energy is small compared to the deformation energy (typically less than 1 %) when
explicit analyses are used to find quasi-static response.
Due to the typical large number of time steps in explicit analyses, the numerical representation of
decimal numbers is important for the stability of the solution. The software options to use high
precision (“double precision”) float are generally preferred.
4.3.3 Solution control for dynamic implicit analysis
A large number of time integration procedures exists (e.g. The Newmark family of methods and the α-
Method). For non-linear analyses they should be used in combination with Newton iterations. As a
rule of thumb the time step should not be larger than 1/10 of the lowest natural period of interest.
The most commonly used integration procedures can be tuned selecting the controlling parameters.
The parameters should in most cases be selected to give an unconditional stable solution.
For the α-method ( HHT method) ref. /26/ the parameters α, β and γ can be selected by the user. The
method is unconditionally stable if:
1 1 1
𝛽 = (1 − 𝛼)2 , 𝛾 = − 𝛼 and − ≤ 𝛼 ≤ 0 (4)
4 2 3
Selecting α less than 0 gives some numerical damping. In order to avoid “noise” from high frequency
modes, parameters that give some numerical damping can be useful. Table 4-1 presents some
combinations of parameters that give unconditional stability.
In case the dynamic effects are not important, the equation system to solve may be reduced to
The selection of element type and formulation is strongly problem dependent. Points to consider are:
• Shell elements or solid elements
• Elements based on constant, linear or higher order shape functions
• Full vs. reduced, vs. hybrid integration formulations
• Number of through thickness integration points(shell)
• Volumetric locking, membrane locking and transverse shear locking
• Hourglass control/artificial strain energy (for reduced integration elements)
In general higher order elements are preferred for accurate stress estimates; elements with simple
shape functions (constant or linear) will require more elements to give the same stress accuracy as
higher order elements. Constant stress elements are not recommended used in the area of interest.
4.5.1 General
The element mesh should be sufficiently detailed to capture the relevant failure modes:
• For ductility evaluations, preferably several elements should be present in the yield zone in order
to have good strain estimates.
• For stability evaluations, sufficient number of elements and degrees of freedom to capture relevant
buckling modes, typically minimum 6 elements per expected half wave should be used.
The element aspect ratio should be according to requirements for the selected element formulation in
the areas of interest.
Care is required in transitioning of mesh density. Abrupt transitioning introduces errors of a numerical
nature.
Load distribution and load type also have an influence on the mesh density. Nodes at which loads are
applied need to be correctly located, and in this situation can drive the mesh design, at least locally.
4.5.2 Mesh refinement study
Often it will be necessary to run mesh sensitivity studies in order to verify that the results from the
analyses are sufficiently accurate.
The analyst should make sure that the element mesh is adequate for representing all relevant failure
modes. In the general case mesh refinement studies may be done by checking that convergence of the
results are obtained e.g. by showing that the results are reasonably stable by rerunning the analysis
with half the element size. See example in section 8.1 .
4.6 Geometry modelling
Geometry models for FE analyses often need to be simplified compared to drawings of the real
structure. Typically small details need to be omitted as they interfere with the goal of having a good
regular element mesh.
The effect the simplifications may have on the result should be evaluated. Typical simplifications
include:
• Cut-outs or local reinforcements are not included
• Eccentricities are not included for beam elements or in thickness transitions in shell models
• Weld material is not included
• Welded parts are modelled as two parts and joined using contact surfaces,
4.7.1 General
The selected material model should at least be able to represent the non-linear behaviour of the
material both for increasing and decreasing loads (unloading). In some cases the material model also
needs to be able to account for reversed loading.
The material model selected needs to be calibrated against empirical data (see 3.3). The basic principle
is that the material model needs to represent the structural behaviour sufficiently for the analysis to be
adequately calibrated against the empirical basis.
4.7.2 Material models for metallic materials
For metallic materials time independent elasto-plastic models are often used. The main components in
such models are:
• A yield surface, defining when plastic strains are generated.
von Mises plasticity is commonly used for metals. The model assumes that the yield surface is
unaffected by the level of hydrostatic stress.
• A hardening model defining how the yield surface changes for plastic strains
Commonly used are isotropic hardening (expanding yield surface) and kinematic hardening
(translating yield surface) or a combination of both.
• A flow rule (flow potential) defining the plastic strain increment a change in stress gives.
The yield surface function is often used as a flow potential (associated flow).
The von Mises yield function is considered suitable for most capacity analyses of steel structures.
The hardening rule is important for analyses with reversed loading; a material model with kinematic
(or combined kinematic/isotropic) hardening rule should be used in such analyses.
Figure 4-2 The von Mises yield surface shown in the σ1-σ2 plane with isotropic (left)
and kinematic (right) hardening models
For small deformations/strains, all strain measures gives similar results. For larger
deformations/strains the strain measure is important, e.g. the Green-Lagrange measure is limited to
“small strains” only. Figure 4-4 shows a comparison of some strain measures. Limitations in the
formulations on the use of the selected element type should always be noted and evaluated for the
intended analysis.
The relationship between Engineering (Nominal) stress and True (Cauchy) stress (up to the point of
necking) is:
The relationship between Engineering (Nominal) strain and True (Logarithmic) strain is:
The stress-strain curve should always be given using the same measure as expected by the software/
element formulation.
As element strain in FE- analyses is an averaged value dependent on the element type and element
size, the reported strain will always depend on the computer model. It is often necessary to re-mesh
and adjust the analysis model after the initial analyses are done in order to have a good model for
strain estimates.
Strain extracted from element integration points are the calculated strain based on element
deformations. Most FE software presents nodal averaged strains graphically. At geometry
intersections the nodal average value may be significantly lower than the element (nodal or integration
point) strain if the intersecting parts are different loaded. When evaluating strain results against
ductility limits, the integration point strains or extrapolated strains from integration points should be
used.
When defining the material curve for the analysis, the following points should be considered:
• Characteristic material data should normally be used, see section 3.3.
• The predicted buckling capacity will depend on the curve shape selected, thus equivalent
imperfection calibration analyses and final analyses should be performed using the same material
curves.
• The extension of the yield zones and predicted stress and strain levels depend on the curve shape
selected. Acceptance criteria should thus be related to the selected material curve, the curve need
not represent the actual material accurately as long as the produced results are to the safe side.
• The stiffness of most steels reduces slightly before the nominal yield stress is reached; in fact yield
stress is often given as the stress corresponding to 0.2% plastic strain
• Some steels have a clear yield plateau; this is more common for mild steels than for high strength
steels.
• One should avoid using constant stress (or strain) sections in the material curves, due to possible
numerical instability issues.
For common offshore steel material qualities the properties shown in Table 4-2, Figure 4-5 and
Figure 4-6 are considered reasonable when ductility (plastic strain) or stability is the acceptance
criteria. The stress-strain values are given using the engineering stress-strain measure. Table 4-3 and
Figure 4-7 show the corresponding True stress-strain values.
The yield and ultimate stress properties should be adjusted for thickness effects according to the
material standard used. The curves should also be adjusted for temperature effects as appropriate.
Table 4-2 Proposed non-linear properties for common offshore steels (Engineering
stress-strain)
S235 S355 S420 S460
E [MPa] 210000 210000 210000 210000
σprop/σyield 0.9 0.9 0.9 0.9
Ep1/E 0.001 0.001 0.001 0.001
σprop [MPa] 211.5 319.5 378.0 414
σyield [MPa] 235 355 420 460
σyield2 [MPa] 238.4 358.4 421.3 461.3
σult [MPa] 320 450 500 530
εp_y1 0.004 0.004 0.004 0.004
εp_y2 0.02 0.02 0.01 0.01
εp_ult 0.2 0.15 0.12 0.1
Ep2/E 0.0022 0.0034 0.0034 0.0036
The proposed material curves in Section 4.7.5 can be used for strain rates up to 0.1 s-1. For impact
loads higher strain rates may be experienced, and the increased strength and reduced ductility may be
considered.
The Cowper-Symonds (CS) model is one model often used to simulate strain rate effects:
1
𝜀̇ 𝑝 (8)
𝜎𝑑𝑦𝑛𝑎𝑚𝑖𝑐 = 𝜎𝑠𝑡𝑎𝑡𝑖𝑐 �1 + � � �
𝐶
The selected model boundary condition needs to represent the real condition in a way that will lead to
results that are accurate or to the safe side.
Often it is difficult to decide what the most “correct” or a conservative boundary condition is. In such
cases sensitivity studies should be performed.
4.9 Load application
Unlike linear elastic analyses, where results from basic load cases can be scaled and added together,
the sequence of load application is important in non-linear analyses. Changing the sequence of load
application may change the end response.
The loads should be applied in the same sequence as they are expected to occur in the simulated
condition/event. E.g. for an offshore structure subjected to both permanent loads (such as gravity and
buoyancy) and environmental loads ( such as wind, waves and current); the permanent loads should be
incrementally applied first to the desired load level, then the environmental load should be
incremented to the target level or collapse.
In some cases the initial load cases (e.g. permanent loads) may contribute positively to the load
carrying capacity for the final load case, in such cases a sensitivity study on the effect of reduced
initial load should be performed.
The analyst needs to evaluate if the loads are conservative (independent of structure deformation) or
non-conservative (follow structure deformation) and model the loads correspondingly.
The number of time/load increments used to reach the target load level may also influence the end
predicted response. Increment sensitivity studies should be performed to ensure that all failure modes
are captured.
4.10 Application of safety factors
Applying load and resistance safety factors in a non-linear analysis can be challenging as application
of safety factors on the capacity model side for one failure mode may influence the capacity of another
failure model. One example of this is yielding vs. column buckling capacity.
In general it is more practical to prepare one capacity model representing the desired characteristic
capacity for all failure modes to be analysed for, and then apply all the safety on the load side,
defining a target load level that accounts for both load and resistance safety. Using this approach, the
same model may be used for both ULS and ALS type of analyses without recalibration of the model:
where Rk is the characteristic resistance found from the analysis, and Sk is the characteristic load
effect.
4.11 Execution of Nonlinear FE analyses, quality control
The analysis should be documented sufficiently detailed to allow for independent verification by a
third party, either based on review of the documentation, or using independent analyses. The
documentation should include description of:
• Purpose of the analysis
• Failure criteria
• Geometry model and reference to drawings used to create the model
• Boundary conditions
• Element types
• Element mesh
• Material models and properties
• Loads and load sequence
• Analysis approach
• Application of safety factors
• Results
• Discussion of results
• Conclusions
Sensitivity studies and other quality control activities performed in connection with the analyses
should also be documented
5.1.1 General
An accurate analysis of tensile failure is demanding as numerous factors affect the problem and
determination of the results from the analysis is highly influenced on how the analysis is carried out.
The recommendations given in this document is not valid for failure that is related to unstable fracture
due to either insufficient material toughness, defects outside fabrication specifications or cracks. In
such cases fracture mechanics methods need to be used.
Table 5-1 Critical strain and net area ratio for uniaxial stress state 1), 2)
Equivalent plastic critical linearized strain 1)
S235 S355 S420 S460
Critical gross yield strain 0.05 0.04 0.03 0.03
Net section ratio 0.94 0.95 0.96 0.97
1) The strain can be calculated as average values over a length of up to 2 times the thickness
2) Any strain due to cold-forming should be subtracted from above values (equivalent strain)
The material model should be according to 4.7.5. For shell and solid models the strain should be based
on average strains calculated on the basis of all elements across the cross-section. The strain should be
based on integration points and linearly extrapolated to the most critical point of the section. The
element mesh should be sufficiently dense to establish accurate strain values or values to the safe side.
Sensitivity analyses may be required to show that the result is sufficient reliable.
5.1.3.3 Strain limits for local tensile failure plane plates (uniaxial stress state)
When the cross-section is weakened by cut-outs or holes that are included in the analysis the local
strain should be checked. The equivalent plastic strains given in Table 5-2 is valid for cases where the
following requirements are followed:
• Maximum element length to be 0.25 of the shortest length of the cut-out or minimum 16
elements around edge of openings
• The strain is taken as the maximum plastic principal strain averaged over the integration points
for the element with the larges strain
• The material model according to Section 4.7.5
• Shear strain due to out-of-plane bending should be checked separately as the average plastic
shear strain across the failure plane extrapolated from the nearest planes of integration points.
Critical plastic out-of-plane shear strain to be according to Table 5-1
Table 5-2 Critical local maximum principal plastic strain for uniaxial stress states 1)
Maximum principal plastic critical strain
S235 S355 S420 S460
Critical local yield strain 0.15 0.12 0.10 0.09
1) Any strain due to cold-forming should be subtracted from above values (equivalent strain)
5.1.3.4 Gross yielding of shell plating in biaxial membrane and bending mode
Gross yielding can be checked in case of biaxial membrane and bending by determining the average
linearized strain in a t*t*t cube at the point of the largest strain. The strains should be checked
separately for strains due to membrane stresses combined with in-plane shear and for shear due to out-
of-plane bending. The strains should be determined by fitting a straight line to the maximum principal
membrane plastic strain based on the integration points and extrapolated to the external corner nodes.
The curve fitting could be made based on the method of least squares. From the linearized strain the
5.2.1 General
Non-linear FE-analyses may imply that the structure is assumed to be loaded beyond proportionality
limits. This means that the structure may be weakened against subsequent load cycles by repeated
yielding leading to a possible cyclic failure. This is called low cycle fatigue and need to be treated
different from how high cycle fatigue checks are carried out.
The fatigue damage due to loads that leads to repeated yielding, i.e. cyclic plastic strains, will be
under-estimated if conventional linear elastic methods, such as those presented in DNV-RP-C203 ref
/25/, are applied. The methodology presented in the following must therefore be applied if repeated
yielding occurs.
Figure 5-1 The true cyclic stress-strain curve for common offshore steel grades
The parameters in Equation (12) are given in Table 5-5 for air and seawater with cathodic protection.
Note that it is allowed to apply parameters for air regardless environment provided that the detail is
crack free and that the loads are subjected during a short time period, e.g. during a storm.
Table 5-5 Data for low cycle fatigue analysis of welded joints
Environment σf' εf'
Air 175 0.095
Seawater with cathodic protection 160 0.060
Figure 5-2 ε-N curves for welded tubular joints in seawater with cathodic protection and air.
Values of the parameters in Equation (13) are given in Table 5-6 for air and seawater with cathodic
protection. Note that it is allowed to apply parameters for air regardless environment provided that the
detail is crack free and that the loads are subjected during a short time period, e.g. during a storm.
Table 5-6 Data for low cycle fatigue analysis of base material
Environment σf' εf'
Air 175 0.091
Seawater with cathodic protection 160 0.057
Figure 5-3 ε-N curve for low cycle fatigue for tubular joint in seawater with cathodic
protection.
For cases where the structure will be loaded by cyclic loads in a way that incremental plasticity may
accumulate and in the end lead to tensile failure or excessive deformations the maximum accumulated
strain need to be checked against the strain values in Table 5-1 for uniaxially loaded plates and
Table 5-3 for biaxially loaded plates.
The criteria for excessive deformations may alternatively be determined on a case by case basis due to
requirements to the structural use or performance. Cases where accumulated strain may need to be
checked can be structure that are repeatedly loaded by impacts in the same direction or functional
loads that change position or angle of attack. Examples of the first are protection structures that are hit
by swinging loads and the latter may be wheel loads on stiffened plate decks.
5.4 Buckling
5.4.1 General
The buckling resistance of a structure or structural part is a function of the structural geometry, the
material properties, the imperfections and the residual stresses present. When the buckling resistance
is determined by use of non-linear methods it is important that all these factors are accounted for in a
way so the resulting resistance meets the requirement to the characteristic resistance or is based on
assumptions to the safe side.
Three different methods for carrying out the analysis are proposed in the following:
a) Linearized approach: Apply the FE-method for assessing the buckling eigenvalues (linear
bifurcation analysis)and determine the ultimate capacity using empirical formulas,
b) Full non-linear analysis using code defined equivalent tolerances and/or residual stresses and
c) Non-linear analysis that is calibrated against code formulations or tests.
Either of these methods can be used to determine the resistance of a structure or part of a structure and
recommendations for their use are given in the following sections.
The proposed methods are valid for ordinary buckling problems that are realistically described by the
FE-analysis. Care should be exercised when analysing complex buckling cases or cases that involve
phenomena like snap through, non-conservative loads, interaction of local and global stability
problems etc.
5.4.2 Determination of buckling resistance by use of linearised buckling values
5.4.2.1 General
If the linearized buckling values (eigenvalues) are calculated for a structure or part of the structure the
buckling resistance can be determined according to the following steps:
σki = kg σRep
(14)
𝑓𝑦 (15)
𝜆=�
𝜎𝑘𝑖
vii) Select empirically based buckling curve to be used based on the sensitivity of the problem with
respect to imperfections, residual stresses and post buckling behaviour. Relevant buckling curves can
be selected from codes, but if not available the following may be used:
Table 5-7 Buckling curves.
Type of buckling κ
1
Column and stiffened plate 2
and plate without 𝜑 + �𝜑 2 − 𝜆
redistribution possibilities ≤ 1.0
2
𝜑 = 0.5 �1 + 𝛼�𝜆 − 0.2 � + 𝜆 � (16)
1,4
1,2
Plate
0,8
Column
Shell
0,6
0,4
0,2
reduced slenderness
Figure 5-4 Examples of buckling curves showing sensitivity for imperfections etc. for
different buckling forms
Empirical based buckling curves are needed to account for the buckling resistance reduction effects
from imperfections, residual stresses and material non-linearity. The effect is illustrated in Figure 5-4.
For all buckling forms the usable buckling resistance is less than the critical stress for reduced
slenderness less than 1.2. Above this value, plates with possibility of redistributing stresses to
longitudinal edges may reach buckling capacities above the critical, column buckling problems will be
less than the critical value, but approach the critical value for large slenderness. Shell buckling is more
sensitive to imperfections and the difference between the buckling capacities that may be exploited in
real shell structures are considerably less than the critical also for large slenderness.
Members will buckle as columns for cross-section classes 1,2 and 3 with exception of tubular sections
exposed to external hydrostatic pressure. For definition of cross-sectional classes see DNV-OS-C101
/13/.
5.4.2.2 Correction for local buckling effects
There may be cases where a reliable finite element representation of local buckling phenomena is not
feasible. This may for instance be torsional buckling of stiffener or local stability of stiffener flange
and web. For such cases the eigenvalue analysis should be carried out without the local buckling
modes represented and the interaction of local and global buckling may be accounted for in a
conservative manner by linear interaction as shown in Equation (18).
1 1 1 (18)
= +
σ ki σ kig σ kil
5.4.3 Buckling resistance from non-linear analysis using code defined equivalent tolerances
The buckling resistance of a structure or part of a structure can be determined by performing non-
linear analyses where the effects of imperfections, residual stresses and material non-linearity is
accounted for by use of a defined material stress-strain relationship and the use of empirically
determined equivalent imperfections. This method is only valid for buckling problem similar to the
cases where the equivalent imperfections are given in Table 5-8.
The material model to be used with the equivalent imperfections is shown in Figure 5-5 or with the
models proposed in 4.7.5.
E/100
1
Stress
Strain
Figure 5-5 Material model for analysis with prescribed equivalent imperfections
It is required that an eigenvalue analysis is carried out to determine the relevant buckling modes.
Usually the pattern from the buckling can be used as the selected pattern for the imperfections, but in
certain cases e.g. when the shape of the buckling load differ from the deflected shape from the actual
loads it may be necessary to investigate also other imperfection patterns.
Figure 5-6 Example of local (left) and global (right) imperfections for stiffened panel
5.4.4 Buckling resistance from non-linear analysis that are calibrated against code
formulations or tests
Buckling resistance can be found by non-linear methods where the effect of imperfections, residual
stresses and material non-linearity is taken care of by use of equivalent imperfections and/or residual
stresses by calibrating the magnitude of the imperfections (and, or the residual stresses) to the
resistance of a known case that with regard to the stability resistance resembles the buckling problem
at hand.
The following procedure based on that an equivalent imperfection is accounting for all effects
necessary to obtain realistic capacities is proposed as follows:
i) Prepare a model that is intended to be used for the analysis.
ii) Perform an eigenvalue analysis to determine relevant buckling modes.
iii) Select the object for calibration and prepare a model using the same element type and mesh density
as intended for the model to be analysed.
iv) Perform eigenvalue analysis of the calibration object and determine the appropriate buckling mode
for the calibration object
vi) Determine the magnitude of the equivalent imperfection that will give the correct resistance for the
calibration object
vii) Define an equivalent imperfection for the most relevant failure mode for the problem under
investigation based on the results from the calibration case
The definition of the equivalent imperfection may in certain cases not be obvious and it will then be
required to check alternative patterns for the equivalent imperfections.
Usually an imperfection pattern according to the most likely of the buckling eigenmodes will be
suitable to be used. Exceptions may be cases where the pattern of the deflected shape from the loads
differ from the shape of the buckling eigenmodes. In cases of doubt several patterns may be needed.
Example of the use of this procedure is included in the Section 8.1.
5.4.5 Strain limits to avoid accurate check of local stability for plates and tubular sections
yielding in compression.
5.4.5.1 General
For cases where compressed parts of the cross-section (as a flange) are experiencing plastic strain in
compression, but one want to avoid an accurate stability analysis of the local buckling effects the
stability can be assumed to be satisfactory if the plastic strain are limited to the values given below.
𝑡 355 (19)
𝜀𝑐𝑟 ≤ �2.7 � � 2 − 0,0016� � 𝑏𝑢𝑡 0 < 𝜀𝑐𝑟 < 0.10
𝑏 𝑓𝑦
𝑡 2 355 (20)
𝜀𝑐𝑟 = �0.29 � � − 0,0016� � 𝑏𝑢𝑡 0 < 𝜀𝑐𝑟 < 0.10
𝑐 𝑓𝑦
𝑡 2 355 (21)
𝜀𝑐𝑟 = �8.5 � � − 0,0016� 𝑏𝑢𝑡 0 < 𝜀𝑐𝑟 < 0.10
𝐷 𝑓𝑦
The strain shall be calculated as plastic strain an may be taken as the average value through a cross-
section of the compressed plated for element length no less than 2 times the plate thickness. Material
properties should be according to 4.7.
For structural parts meeting requirements to sectional class 3 or 4 no plastic strain due to compressive
stresses can be allowed without an accurate buckling analysis.
For definition of sectional classes see DNV-OS-C101 /13/.
5.5 Repeated buckling
For cases where buckling of parts of the structure may occur before the total capacity of the entire
structure is reached, it is necessary to investigate if the buckling may cause reduced capacity against
cyclic loads. When significant cyclic loads are present one should limit the capacity to the load level
that corresponds to the first incident of buckling or a cyclic check needs to be carried out. See 5.2.2 for
determination of cyclic loads.
For cyclic loads following an extreme wave or wind load, it is regarded acceptable to disregard failure
due to repeated buckling of the following cases:
• Buckling of the individual plates in a stiffened plate structure if the plate to span ration is less
than 120.
• Member buckling if all parts of the cross-section meet requirements to Cross-sectional class 1
and the reduced member slenderness as a column is above 0.5.
Failure due to low cycle fatigue according to 5.2 needs to be checked also for these cases.
It should be noted that structural parts that is yielding in tension may buckle when unloaded and that if
cyclic loads leads to yielding in tension one need to check against buckling through the entire
dimensioning load cycle.
6 BIBLIOGRAPHY
/1/ ISO2394, General principles on reliability for structures, Second edition 1998-06-01
/2/ API RP 2A Recommended Practice for Planning, Designing and Constructing Fixed Offshore
Platforms – Working Stress Design, Errata and supplement 3 October 2007
/3/ EN 1990, Eurocode - Basis of structural design, April 2002
/4/ EN 1993-1-1 Eurocode 3 Design of steel structures. Part 1-1 General rules and rules for
buildings
/5/ EN 1993-1-5, Eurocode 3 - Design of steel structures - Part 1-5: Plated structural elements,
October 2006
/6/ EN 1993-1-6, Eurocode 3 - Design of steel structures - Part 1-6: Strength and Stability of Shell
Structures, February 2007
/7/ EN 1993-1-8, Eurocode 3 - Design of steel structures - Part 1-8: Design of Joints,
2005/AC:2009
/8/ ISO 19900 Petroleum and natural gas industries – General requirements for offshore structures.
First edition 2002-12-01
/9/ ISO 19902 Petroleum and natural gas industries – Fixed steel offshore structures, First edition
2007-12-01
/10/ Norsok Standard N-001, Integrity of offshore structures, Edition 7, June 2010
/11/ Norsok Standard N-004, Design of steel structures, Revision 3 October 2004
/12/ Norsok Standard N-006, Assessment of structural integrity for existing offshore load-bearing
structures, Edition 1, March 2009
/13/ DNV-OS-C101, Design of Offshore Steel Structures, General (LRFD Method) April 2010
/16/ ECCS publication No. 125, Buckling of Steel Shells. European Design Recommendations, 5th
Edition, J.M. Rotter and H. Smith Editors.
/17/ DNV-RP-F110 Global Buckling of Submarine Pipelines Structural Design due to High
Temperature/High Pressure October 2007
/18/ DNV-SINTEF-BOMEL: ULTIGUIDE, Best practice for use of non-linear analysis methods in
documentation of ultimate limit state for jacket type offshore structures, April 1999.
/19/ Skallerud, Amdahl: Nonlinear analyses of offshore structures, Research studies press ltd., 2002
( ISBN 0-86380-258—3)
/20/ Corrocean ASA: Design of offshore facilities to resists gas explosion hazards. Engineering
handbook. Oslo 2001.
7 COMMENTARY
7.1 Comments to 4.1 General
The element model selected for analysis need to represent the structure so that any simplifications are
leading to results to the safe side. This is especially important for the selection of boundary conditions
and the representation of the load. The analyst needs to assess the possibility that simplification may
lead to an overrepresentation of the resistance. An example may be the representation of neighbouring
elements that also are subjected to buckling. In the case that the stiffness of the adjoining structure is
uncertain it is recommended to use boundary condition corresponding to simple support. If there are
uncertainties with respect to simplification in load it is recommended to vary the load pattern and
perform alternative analyses to check the effect.
The requirements to characteristic resistance in other codes for offshore structures like ISO 19902 /9/
is similar and the analysis carried out according to the recommendations in this RP is expected to fulfil
the requirements also in this code.
7.2 Comments to 4.4 Selection of Elements
Guidance on selection of suitable elements for non-linear analysis can be found in text books e.g. /31/.
7.3 Comments to 5.1.1 General
There are several models describing the local phenomenon of tensile failure. Common for most of
these is that the strain and stress state during the entire loading sequence until failure is considered
important for describing the damage process properly. Unfortunately, high degree of complexity is a
common feature of many of the models, and the theoretical and practical knowledge required to
perform a finite element analysis based on these criteria is above the level that can be expected of a
common engineer.
7.5 Comment to 5.2.2 Determination of cyclic loads
The check against cyclic failure should be carried out with the use of a dimensioning load history that
has the prescribed probability of occurrence as required for a single extreme load. For environmental
loads like wave and wind it should be established a dimensioning storm that the structure should
survive. It would be in line with check for other failure modes to check the structure for one single
storm from each of the critical directions, but without adding the calculated damage from different
directions.
The load-history for the remaining waves in a 10 000 year dimensioning storm investigated for
southern North Sea conditions have been found to have a maximum value equal to 0,93 of the
dimensioning wave, a duration of 6 h and a Weibull shape parameter of 2,0. This applies for check of
failure modes where the entire storm will be relevant, such as crack growth.
When checking failure modes where only the remaining waves after the dimensioning wave (e.g.
buckling) need to be accounted for, a value of 0,9 of the dimensioning wave may be used ref /24/.
The load-history for the remaining waves in a 100 year dimensioning storm investigated for southern
North Sea conditions have been found to have a maximum value equal to 0,95 of the dimensioning
wave, a duration of 6 h and a Weibull shape parameter of 2,0. The largest remaining waves after the
dimensioning wave (e.g. for cases like buckling) the largest wave is found as 0.92 of the dimensioning
wave.
7.6 Comment to 5.2.3 Cyclic stress strain curves
Note that the cyclic stress strain curves are only intended for low cycle fatigue analysis. The cyclic
stress strain curve is different from the monotonic curve and the stress value is often below due to
cyclic softening. This is the case for the material presented in this RP, cf. Figure 4-7 and Figure 5-1.
Laboratory test results presented in references /27/-/30/ make basis for the established ε-N curve for
welded joints. The proposed mean and design curve for air along with the laboratory test data is
presented in Figure 7-1. Note that some of the results presented in the figure are not obtained directly
from the referred articles. In some cases further analysis and interpretation was needed to obtain the
data on a proper format.
The mean curve is established based on judgement. The results reported by Weigans and Berman ref
/28/ are obtained from testing of dog-bone specimens cut out from a butt welded plate. These results
have therefore been weighted less than results from ref /27/ and ref /29/ which is based on full scale
testing of tubular joints. The fatigue test results presented in ref /30/ are from pipes with wall
thicknesses of less than 10mm. The fatigue strength of welded joints is to some extent dependent on
the wall thickness and since the thickness of structural elements normally is significantly larger than
this the results have been weighted less.
Because the fatigue test data comes from several different sources it was not found reasonable to
establish the standard deviation from a regression analysis. Instead, a standard deviation of 0.2 in log
N scale is assumed for constructing the design curve in air. A standard deviation of 0.2 is identical to
Figure 7-1 Mean and design curve for welded joints along with laboratory test results.
When a structure is loaded beyond linear limits the response for subsequent cycles will be changed. It
is therefore necessary to investigate the behaviour through the full cycles also for the next cycles. See
e.g. /19/ for more guidance.
7.9 Comments to 5.4.1. General
The modelling of geometrical imperfections, out of straightness etc. is central for achieving a credible
and safe estimate of the buckling and ultimate strength limit. The less redundant the structure is the
more important it will be to model the geometrical deviations from perfect form in a consistent way
using the eigen-mode, postbuckling shapes, combinations thereof or similar. For redundant structures
the sensitivity of the ultimate load bearing capacity to the size of the geometrical imperfections will be
negligible. For such cases the triggering of the governing modes rather than accounting for actual
tolerance size will be most important for the analyses.
The strain limits for plates are established from analysis of flanges meeting rotational capacities
according to cross-section class 1 and 2 and by comparison with tests. See /33/ and /34/. Strain limits
are also compared with recommendations given in the DNV pipeline standard /35/.
8 EXAMPLES
8.1 Convergence test of linearized buckling of frame corner
A frame of beams with I-section is analysed. The frame with boundary conditions is shown in
Figure 8-1. The loading is applied as a displacement of the web of one end of the frame, 𝑢2,𝑎𝑝𝑝𝑙𝑖𝑒𝑑 =
−0.01 m. Three different mesh densities and two element types are included in a convergence study,
to ensure a sufficiently refined mesh. See Figure 8-3. The element types used are 4 node rectangular
shell elements and 8 node rectangular shell elements.
The analyses are performed using the FEM-software ABAQUS.
Figure 8-3 Top: coarse mesh. Middle: fine mesh. Bottom: very fine mesh
The loading is applied as displacement on the web of one end of the frame, as shown in Figure 8-2.
Hence, the eigenvalue defines the displacement corresponding to linearized buckling.
A convergence study is performed by analysing 6 cases and the resulting buckling displacements are
listed in Table 8-2. From these results all combinations of mesh size and element type except the
coarse 4 node combination seems to be sufficiently refined. However, the stress results wanted are
also highly dependent on the mesh refinement, and a fine mesh in the area where high stress values are
reached is preferable. An analysis using the very fine mesh is time consuming, hence the mesh size
and element type combination chosen is the 4 node elements with fine meshing.
Table 8-2 Convergence study of frame
Case Mesh size Element type Linearized buckling displacement [m]
number
1 Coarse 4-node 0.0653
2 Fine 4-node 0.0624
3 Very fine 4-node 0.0618
4 Coarse 8-node 0.0616
5 Fine 8-node 0.0615
6 Very fine 8-node 0.0615
In summary the convergence test has shown that case number 2 and case 4 will produce sufficient
accurate results of the linearized buckling value. Case 2 is preferred as the analysis is more efficient
compared to case 4. The increased mesh refinement of case 3, 5 and 6 will not improve the accuracy
for the actual problem.
Figure 8-5 Stress distribution of compressive stress from linear analysis (min principal
membrane) (Min. In-plane principal stress)
𝑓𝑌
𝜆̅ = � = 0.76
𝜎𝑘𝑖
𝜅𝑓𝑦 𝑆𝑅𝑒𝑝
𝑅𝑑 =
𝛾𝑀 𝜎𝑅𝑒𝑝
A non-linear analysis (using the arc-length method) is performed, where the effects of imperfections,
residual stresses and material non-linearity is accounted for by use of a defined material stress-strain
relationship and the use of empirically determined equivalent imperfections. The shape of the
governing buckling mode is taken as the lowest buckling mode as shown in Figure 8-6, and is used as
the pattern for the equivalent imperfection. The magnitude of the equivalent imperfection 𝛿 is
calculated using the tolerances given in Table 5-8.
The analysed frame can be considered equivalent to a component of longitudinal stiffener or flange
outstand, hence the magnitude is given as
𝛿 = 0.02 rad = 0.02𝑐
where c is half the width of the flange. Two values of c are analysed, the largest c; 𝑐𝑎 = 𝑎, where
𝑎 = 0.975m is the distance between where the webs cross in the corner of the frame and the midpoint
𝑏
𝑎+
of the flange curvature, and an average c; 𝑐𝑎𝑣𝑔 = 2
, where 𝑏 = 0.5 m is the width of the flange
2
𝛿𝑎 = 0.0195
𝛿𝑎𝑣𝑔 = 0.0122
Figure 8-8 Stress distribution for non-linear analysis with largest initial imperfection at
maximum applied force
8.3.2 Results
Figure 8-9 displays the force-displacement curves for the displaced end of the frame for the linear
analysis and the force-displacement corresponding to the critical buckling stress where imperfections
are taken into consideration as calculated in Section 8.2, and from the non-linear analyses.
Figure 8-9 Force-displacement from non-linear analyses, linear analysis and the
calculated critical value
Figure 8-10 Geometry of conical transition (on top) and calibration object (bottom),
dimensions in mm
Figure 8-11 Left: Load and boundary conditions. Right: Element mesh
8.4.3 Step iii: Select object for calibration and prepare model
The calibration object is selected as a cylinder. The diameter and wall thickness are equal to the lower
cylindrical part of the conical transition, while the length is chosen as 2/3 of the conical transition
length (lower part, conical part and a part of the top part). The load and boundary conditions, element
type and mesh density used is the same as for the model of the conical transition, see Figure 8-13.
Figure 8-13 Left: Load and boundary conditions. Right: Element mesh
8.4.4 Step iv: Determine the appropriate buckling mode for the calibration object
Eigenvalue analysis is performed to find the buckling modes for the calibration object. These buckling
modes are compared to the buckling modes found for the conical transition and a mode with similar
pattern is selected. Figure 8-14 shows the first cylinder buckling mode. This shows a similar pattern to
the buckling mode of the conical transition Figure 8-12, hence this is determined to be an appropriate
buckling mode.
8.4.6 Step vi: Perform non-linear analysis of the model with imperfections
A non-linear analysis of the conical transition with imperfections is performed. The load and boundary
conditions remain the same, and the material model and magnitude of the calibrated imperfection from
Step v is used. The load proportionality factor for this case is shown in Figure 8-15. The maximum
load proportionality factor is 𝐿𝑃𝐹𝑚𝑎𝑥 = 0.936489. Thus the buckling capacity of the conical
transition subjected to the given load combination is; hydrostatic pressure 𝑝 = 0.95MPa and an axial
tension 𝑁𝑦 = 54.7MN. Figure 8-16 shows the von Mises stress at maximum load on the deformed
conical transition.
Figure 8-15 Load proportionality factor for conical transition with initial imperfection
Figure 8-16 Deflected shape showing von Mises stress at maximum load deformations
scaled with a factor of 10
8.5 Example: Low cycle fatigue analysis of tubular joint subjected to out of plane
loading.
This example presents a low cycle fatigue analysis of a tubular T-joint subjected to an out-of-plane
fully reversible load of ± 50 kN. The objective of the analysis is to estimate the design life based on
It is assumed that the cyclic stress-strain behaviour is well described by the Ramberg-Osgood
relationship:
1
𝜎 𝜎 𝑛′
𝜀 = + � ′�
𝐸 𝐾
The values for the Ramberg-Osgood parameters are presented in Table 8-6 for the chord and the brace.
Table 8-6 Ramberg-Osgood parameters
K’ n’
[MPa]
Chord 731.7 0.096
Brace 699.5 0.108
In order to obtain the cyclic strains a finite element analysis was carried out using the FEM-software
ABAQUS a 8-node shell element (S8R) model was established as shown in Figure 8-18.
The chord was constrained at each end for all translational and rotational degrees of freedom. The out-
of-plane load was applied by means of a reference point located at the cross-section centre of the brace
end. This reference point is connected to the circumference of the brace end by means of kinematic
coupling. The load was applied using two steps as illustrated in Figure 8-19.
Figure 8-20 (a) shows an overview of the finite element mesh. Figure 8-20 (b) shows a close-up of the
brace-chord intersection area. The finite element mesh in the hotspot region is established such that the
strain at 0.5t and 1.5t at both the brace and chord side easily could be obtained at the element mid-
node.
Figure 8-20 Left: (a) Meshed model, Right: (b) Close-up of brace-chord intersection
area
Figure 8-21 shows the principal strain range due to the out-of-plane cyclic loading. Based on the 1st
principal strain range at 0.5t and 1.5t the hotspot strain range is calculated. The hotspot strain range is
calculated both by using the mid-node values and by extrapolating from the integration points and is
presented in Table 8-7.
Table 8-7 Hotspot strain range
𝚫𝜺𝒉𝒔
Nodal value 0.0053
Extrapolated from integration points 0.0051
Air environment is assumed. Hence, the characteristic design life due to the cyclic loading is obtained
by solving the following equation, ref. Section 5.2.4:
∆εhs σ′f
= (2N)−0.1 + ε′f (2N)−0.5
2 E
→ 𝑁 = 642/693 cycles
8.6 Example: Low cycle fatigue analysis of plate with circular hole.
In this example a low cycle fatigue analysis of a plate with a circular hole subjected to a far field stress
range of 275MPa is presented. The objective of the analysis is to estimate the design life based on the
The cyclic equivalent strain is obtained by performing a finite element analysis with the FEM-software
ABAQUS. A 8-node shell element (S8R) model is established as shown in Figure 8-23. The material
modelling is according to specifications in Section 4.7.5. The cyclic load is applied by specifying two
load steps as illustrated in Figure 8-24. The far field stress of 275MPa is obtained by means of the
‘shell edge load’ option in ABAQUS.
8.7 Gross yielding of shell plating in biaxial membrane and bending mode
The current example presents the yield check of a plate with attached lifting ear. The objective is to
determine the allowable vertical load according to the recommendations in this recommended practice.
The material grade of the plate and lifting ear is S355.
The geometry of the lifting ear is only partly modelled in order to simplify the finite element analysis,
see Figure 8-26. The load is applied by means of a reference point that is attached to the shaded area in
the figure. All four plate edges are assumed to be fixed, i.e. zero displacement and rotation.
The loading and boundary conditions will results in a biaxial stress/strain state. Hence, the criterion
presented in Section5.1.3.4 is applied for assessing the integrity of the connection.
The yield check will be carried out based on stress and strain output of a finite element analysis carried
out in ABAQUS. 20-noded solid elements with reduced integration (C3D20R) are used. The finite
element size is between 16mm and 20mm which correspond to the thicknesses of the plate and lifting
ear.
Figure 8-27 Contour colour plot of the plastic shear strain component PE12 (𝜺𝑷𝒙𝒚 ). The
location of the cube is indicated by arrow.
F
E
G
H A
C
D
The failure plane is considered to be the element surface in extension of the lifting ear (surface
BCGF). The average plastic shear strain at the failure plane is obtained by the following procedure:
𝑃
1. Establish plastic shear strain component 𝜀𝑥𝑦 in all eight (Gauss) integration points.
2. Average the plastic shear strain component in the two integration point planes parallel to the
failure plane (planes 2367 and 1458). Numbering of the corner points of the cube and the
integration points are shown in Figure 8-29.
3. Linearly extrapolate the averaged plastic shear strain values to the failure plane.
F E
G H
6 5
7 8
2 1
4
3
B A
C D
Figure 8-29 Sketch of the solid element along with integration point and corner
numbering.
3 4
C D
Based on the finite element analysis and subsequent post-processing of results the critical load was
determined between 810 and 820kN. The results are presented in Table 8-8.
Table 8-8 Analysis results.
Location Average plastic shear strain
810kN 820kN
-2
Surface BCGF -3.5·10 -4.1·10-2
The design resistance will be found as the characteristic resistance divided with the appropriate
material factor.
8.8 Example: Strain limits for tensile failure due to gross yielding of plane plates
(uniaxial stress state)
8.8.1 T-section cantilever beam
Gross yielding check of a T-section cantilever beam, subjected to axial and shear force and moment
loading, is presented in this example. The finite element software ABAQUS is used to perform the
analyses.
The geometry and boundary conditions of the beam are shown in Error! Reference source not
found.. Loading is applied to a reference point coinciding with the neutral axis of the beam cross
section, using kinematic coupling between cross section and reference point. The beam is modelled
using 4-node shell elements with reduced integration (S4R) with mesh size of 16 mm x 16 mm.
Material grade is S355, modelled according to Section 5.1.3.2.
The magnitude of the applied forces and moments are given by axial force 𝑁𝑥 , shear force 𝑃𝑦 =
−0.15𝑁𝑥 and bending moment 𝑀𝑧 = −0.45𝑁𝑥 .
The loading and boundary conditions results in a stress state dominated by uniaxial stress. Hence, the
criterion presented in Section 5.1.3.2 is applied for assessing the beam.
According to the criterion presented in section 5.1.3.2, the strain should be calculated as the linearized
maximum principal plastic strain along the likely failure line and checked against the limit for the
critical strain. The limit is a critical gross yield strain of 0.04 for this example.
Error! Reference source not found. shows a contour plot of the maximum principal plastic strain
and the chosen failure line, the 3rd element column from the clamped end. For the chosen failure line
the maximum principal plastic strain is obtained from integration points and linearized using the
method of least squares.
Based on the finite element results and the linearization the critical load is determined between
𝑁𝑥 = 489 kN and 𝑁𝑥 = 500kN. The maximum principal plastic strain distribution and corresponding
linearized distribution for load level 𝑁𝑥 = 489kN are shown in Error! Reference source not found.
and analysis results are shown in Table 8-9.
Table 8-9 Analysis results.
𝑵𝒙 Maximum linearized
[kN] principal plastic strain
489 3.9·10-2
500 4.2·10-2
Figure 8-32 Maximum principal plastic strain contour plot, with chosen failure line
highlighted
Figure 8-33 Maximum principal plastic strain and linearized maximum principal plastic
strain distributions for web
The design resistance will be found as the characteristic resistance divided with the appropriate
material factor.
Figure 8-34 Geometry and boundary conditions for cantilever beam with notch
For the gross yielding two likely failure lines were chosen; one at mid-notch and one at the notch
corner displaying the highest strain values, see Figure 8-35. The maximum principal plastic strain is
obtained from integration points and linearized using the method of least squares. Both failure lines
must comply with the criterion of an allowable maximum principal plastic linearized strain of 0.04. In
Figure 8-35 Maximum principal plastic strain contour plot, with chosen failure lines
highlighted
Based on the finite element results and the linearization the critical load is determined between
𝑁𝑥 = 310 kN and 𝑁𝑥 = 320kN. The maximum principal plastic strain distributions and corresponding
linearized distributions for both failure lines at load level 𝑁𝑥 = 310kN are shown in Figure 8-36 and
analysis results are shown in Table 8-10.
Table 8-10 Analysis results.
The design resistance will be found as the characteristic resistance divided with the appropriate
material factor.
Figure 8-36 Maximum principal plastic strain and linearized maximum principal plastic
strain distributions for web
The boundary conditions and applied moment and loads (Fx, Fy and Fz) are shown in Figure 8-23. The
magnitude of the forces are given by Fy = 0.15·Fx and Mz = 0.45·Fx. The check is carried out for a
load situation where Fx = 385kN.
Figure 8-23 Illustration of the boundary condition and the load directions.
The yield checks are carried out using non-linear finite element analysis with ABAQUS. The element
type used in the analysis is 8-node shell element (S8R). The hole has been meshed with 16 elements
around the circumference with a nearly quadratic element shape. The global mesh of the component
has an approximate size of 75mm x 75mm. The global and local mesh near the hole is shown in
Figure 8-37.
The distribution of the plastic strain is shown on figure 8-26. The gross yield criterion presented in
Section 5.1.3.2 is based on the linearized 1st principal plastic strain through the tensile part of the cross
section (assessment line). The basis for the linearized strain is shown in Figure 8-38. The strain value
is 6.3% which is above the criterion of 4%.
Figure 8-38 Linearized 1st principal strain along the assessment line.
The local yield criterion is based on the element with the highest mean 1st principal plastic strain value
obtained from the four integration point as specified in Section 5.1.3.3. A values of 8.8% is assessed
which is below the criterion of 12%.