Professional Documents
Culture Documents
Review of Ultimate Strength of Tubular Framed Structures (Oth365)
Review of Ultimate Strength of Tubular Framed Structures (Oth365)
Authors
HSE BOOKS
ISBN 0-7176-1040-3
SUMMARY
1. INTRODUCTION
1 . 1 Background
1.2 Objectives and Scope of the Review
1.3 Development of the Review
3. EXPERIMENTAL INVESTIGATIONS
3.1 Background to Test Programmes
3.2 Comparison of Results
3.2.1 Single versus two-bay plane frames
3.2.2 Role of redundant members
3.2.3 K versus X bracing
3.2.4 Effective length factors
3.2.5 Joint versus member failures
3.2.6
3.2.7 Initial imperfections and the effects of scale
3.2.8 Comparison with jacket structures
3.2.9 Materials
3.2.10 Conclusion
iii
Page
5. ANALYTICAL INVESTIGATIONS
5.1 Background to Analyses
5.1.1 Simple 2D frame analyses
5.1.2 Idealised 3D jacket analyses
5.1.3 Structural jacket analytical investigations
5.1.4 Jacket loading investigations
5.1.5 Jacket hindcasting calculations
5.1.6 Reliability analyses
5.2 Comparison of Results
5.2.1 Quantification of RSR
5.2.2 Bracing configuration
5.2.3 Joint behaviour
5.2.4
5.2.5 Foundation modelling
5.2.6 The role of comparative analysis
6. DISCUSSION
6.1 Summary of Review
6.1.1 Experimental results
6.1.2 Numerical results
6.2 Reserve Strength of Frames
6.2.1 Background to evaluating reserve strength
6.2.2 Redundancy beyond first member failure
6.2.3 Alternative loadpaths and sources of reserve
6.2.4 Relation between 2D and 3D structures
6.3 Reserve Strength Considerations for Offshore Jacket Structures
6.3.1 Complexity of 3D jacket structures
6.3.2 Modelling of jacket loads
Influences on RSR
6.3.4 Role of tubular joint failures
6.3.5 Role of foundation failure
6.3.6 Cyclic loading effects
6.3.7 Accounting for damage
6.3.8 Target system reserve
6.4 Calculation of Reserve Strength
6.4.1 Analytical tools for pushover analysis
6.4.2 Validation
REFERENCES
SUMMARY
This review of the ultimate strength of tubular framed structures has been prepared for the
Health and Safety Executive (HSE) by Billington Osborne-Moss Engineering Limited
(BOMEL). A numerical capability to predict the nonlinear response of jacket structures
has been developed over the last decade in parallel with experimental investigations. It is
now being applied to assure the continued integrity of installations beyond the design event
in circumstances of extreme environmental loading or damage. A recent investigation has
confirmed that an extreme event static pushover analysis generally suffices to demonstrate
a structure's resistance to the cyclic loading of the full storm.
This report draws together the results from published investigations and identifies key
factors contributing to system reserve. It is shown that bracing configurations and relative
member properties are important influences. From the work presented, it is demonstrated
that many jacket analyses embody simplifying assumptions, and features such as loading
asymmetry, joint nonlinearity, foundation interactions, global deflection criteria etc, are
neglected. Specific examples highlighted in the review illustrate their potential importance
and systematic sensitivity evaluations are therefore recommended.
Differences in the definition of reserve strength ratio are noted, underlining difficulties in
drawing comparisons between structures. Nevertheless, jacket examples are cited where
first failure precipitates global collapse. Other structures are sufficiently redundant to
sustain loads well in excess of the design value, with collapse occurring only after a
sequence of component failures under increasing load.
The facilities of specific software programs are compared. Analyses using different
programs are shown not always to give consistent results and discrepancies in terms of
capacity and failure mode for the same jacket structure are found. Further benchmarking
and detailed comparison of software predictions is recommended.
This review was completed in 1993. Since that time and prior to publication of the review
in 1995 a number of important developments have taken place in relation to both the
understanding and the application of ultimate system strength technology.
A supplementary review in Appendix A brings the document up to date reflecting the
insight to frame behaviour derived from Hurricane analyses, the
experience of recent benchmarking activities and the acceptance of ultimate strength
analyses in API RP 2A for the assessment of existing offshore structures.
Note
The illustrations, provided by BOMEL,in this report are only intended to be indicative of
actions that have been taken and not to be clear representations of the subject matter.
1. INTRODUCTION
1 BACKGROUND
The design of jacket structures is generally based on the expected response of components
to the applied loads anticipated. There is uncertainty on both the loading and resistance
sides of this equation so characteristic values are derived from the available data.
Furthermore, safety factors are introduced explicitly to ensure that an 'adequate' safety
margin exists.
Simplifying assumptions are inherent in the derivation of component forces from global
loads. An elastic frame analysis is performed, typically with elements rigidly connected.
Components are sized to ensure that the acting loads do not exceed the allowable values
designated by the codes for each component. Any potential of the structure to yield and
redistribute loads is neglected, giving an inherent 'reserve' capacity beyond the design event
(typically the year return period storm wave). The risk of exceptional loading beyond
the 100 year event is not negligible however and modem codes (eg. ECCS and NPD) are
taking account of 10,000 year loadings but with plastic responses permitted. In November
1993 an API preliminary draft for RP 2A-WSD Section 17.0 for the assessment of existing
platforms was circulated, in which a sequence of analysis from screening, through design
level to ultimate strength assessment is advocated to demonstrate structural adequacy. At
the ultimate strength level it is proposed that 'a platform may be assessed using inelastic,
static pushover analysis' [see also Appendix A].
This review is concerned primarily with the reserve strength of jacket structures as
evaluated in pushover analysis. The frame action and system redundancy are implicit
sources of reserve strength which are not generally controlled or quantified in design.
Similarly, conservatism embodied within codes, material yield strengths exceeding the
minimum criteria specified, component limit states less onerous than ultimate strength and
overdesign for non-structural requirements, may be considered as implicit sources of
reserve. By contrast, overdesign by exceeding minimum requirements or by conservative
combinations of loads are sources of explicit reserve and can be controlled by the designer.
Lloyd and (1984) present a discussion of these various sources of reserve and
residual strength but the focus of this review is on the important contribution of frame
behaviour. Marshal1(1979) demonstrates that this difference between elastic single element
behaviour and ultimate strength system behaviour is a major source of reserve strength.
Marshal1 and Bea (1976) demonstrated that a reserve strength factor of the order of 2 on
the design capacity may be found in offshore structures. Kallaby and (1975)
published one of the first applications of inelastic analysis to demonstrate the energy
absorption capacity of the Maui A platform under earthquake loading.
Reserve strength should not be soley considered as overdesign of structures, rather it is
required to cope with loads which have not been foreseen in the design process or loads
which cannot be economically designed for on an elastic basis (eg. seismic or accidental
loads). The risks of these are not negligible and whilst traditional elastic design approaches
might preclude economic structural solutions for all conceivable loads, it is essential to
demonstrate that extreme events can be sustained without endangering human life or the
environment.
It is also important that a structure can sustain damage without collapse, ie. that it has
sufficient remaining or 'residual' strength. Such damage may result from extreme
overloading of the structure as a whole or from localised damage (eg. from impact). If
alternative load paths exist, the forces may be redistributed safely.
These requirements are not stipulated in quantitative terms within design codes although,
as noted above, traditional design practices have embodied inherent reserves. The attention
to safety in the post Cullen era has underlined the need to consider hazards such as extreme
environmental conditions or accidental loading scenarios which present a significant risk
to structural survival but which may not have been considered in the traditional design
process. Therefore there has been the requirement to develop an understanding and the
corresponding analytical tools to be able to predict system reserves beyond individual
component failure capacities, in order to demonstrate integrity in the event of such extreme
loading scenarios occurring.
Trends for lighter, liftable jackets and new concepts for deeper waters provide additional
impetus to the study. Fewer members in the splash zone may increase the risk to topsides
safety in the event of impact, and the deletion of members with low elastic utilisations to
save weight reduces the capacity for redistribution along alternative load paths.
Comparative calculations of reserve capacity for different structural configurations can help
ensure that levels of reserve strength and safety embodied within older designs are
maintained.
Reserve strength calculations may therefore be required in the course of the service life or
may be used to optimise a configuration or compare different concepts at the design stage
to ensure efficient structural forms are adopted.
However, the ultimate strength of a structure is not simple to calculate. It depends on the
nonlinear responses of components within a frame and the interaction between those
components. Figure 1.1 illustrates the three primary bracing types (a, b, d) used alone or
in combination in jacket structures. The presence of alternative load paths within a panel
ensures that (d) the X-bracing offers greater reserves than either (b) the K-bracing, or (a)
the single diagonal bracing. However, the degree of reserve depends on the slenderness
of the braces and redundancy throughout the structure. The hybrid structure (c) is not
therefore considered satisfactory in API RP 2A whereas for (e) failure of one
member could be tolerated without structural collapse. Furthermore, the structural reserve
depends on alternative load paths through other panels within the frame, (Figure as
well as on three-dimensional framing between planes (Figure 1.3).
Linear analysis of the first structure in Figure 1.2 would show there to be negligible load
in the horizontals between the panels. However, in the event of damage (diagonal braces
removed), these horizontals would be essential for maintaining framing action through the
structure. The idealised structures in Figure 1.3 were used by Lloyd (1982) to determine
the minimum structural weight to achieve a desired residual capacity beyond primary
member failure. A simple linear programming technique was adopted. Elastic analysis
again showed the face frame horizontals and diagonal plan framing in A to be redundant
for the applied loading regime, but in the event of a brace 'failure' these members provide
important alternative load paths to distribute the forces efficiently down through the
structure.
Since the 1970s experimental programmes have been implemented to provide data on the
collapse behaviour of frames (see Section 3). In addition to revealing the response
characteristics, the results enable reserve strength to be quantified and provide physical data
against which nonlinear software can be verified. Indeed, in parallel with the experimental
work, a number of 'pushover' analysis programs have been developed, embodying not only
material nonlinearity but also large displacement behaviour inherent in structural collapse
(see Section 4). These programs have been applied to a number of frames, representative
of jacket structures. In some instances the full ultimate response has been evaluated; in
others, a simplified approach in which 'damaged' members are removed has been taken to
evaluate the residual capacity (see Section 5).
The investigations (particularly numerical) have often been motivated by specific problems
in the field. Nevertheless, sufficient results now exist for the findings to be drawn together
to start to give an overview and comparison of the reserve strengths of different structural
forms. Reserve strength is an important yardstick of safety and the results of this work will
be useful in both the design of new structures and requalificationand assessment of existing
installations.
It is on this basis that the Health Safety Executive commissioned Billington
Moss Engineering Limited (BOMEL) to undertake the present review of the reserve
strength of framed structures. The principal research effort in the UK in relation to reserve
and residual strength has been undertaken within the Joint Industry Funded Tubular Frames
Project, first at the Steel Construction Institute (Phase I, 1990) and then by Billington
Osborne-Moss Engineering Limited (Phase 1992) to whom the project was transferred.
The project, described more fully in Sections 3, 4 and 5, encompassed collapse tests on
large scale tubular frames as well as the development of advanced nonlinear software for
the pushover analysis of 2D and 3D frames and jackets. This review presents the test
results and places them in the context of other research findings.
Figure 1
Principal bracing configurations-adoptedin offshore jacket structures
'NO LOAD IN
THESE MEMBERS
figure 1.2
Alternative load paths through bracing
figure 1.3
Alternative plan bracing to distribute loads in
2. DEFINITIONS AND PRACTICAL
CONSIDERATIONS
In assessing the ability of a structure to withstand loads in excess of the design load or to
sustain loading in the damaged state, some measure of this ability is required. Terms such
as reserve and residual strength, and redundancy are used and it is appropriate that this
review should begin with a clear definition of these terms and their usage.
Eqn 2.5
2.3 REDUNDANCY
Fixed offshore structures generally have a multiplicity of load paths such that failure of a
single member does not necessarily lead to catastrophic structural collapse. This is
attributable to the 'redundancy' of the system but it is demonstrated below that careful
definition of the term is required.
In conventional deterministic structural engineering, redundancy is generally equated to the
degree of indeterminacy, ie. the number of unknown internal member forces in excess of
the number of degrees of freedom of the system. However, this definition is not
satisfactory for evaluating the ability of a structure to withstand overloads in the intact or
damaged condition. It does not account for the existence of a weak link in an otherwise
highly redundant system or the distribution of redundancy (or under utilisation) throughout
the system. See Figure 1 for an example of such a 'weakly redundant' system. Lloyd
and (1984) suggest that, in practice, each member should systematically be
removed so that the consequences, in terms of the remaining capacity beyond which
progressive collapse occurs, can be evaluated. In this way the concept of residual strength
was developed. They present the hierarchy reproduced in Table 2.1 to demonstrate the
gradation in redundancy that can be afforded by different members. Such an approach can
be used as a basis for sizing members to give adequate redundancy.
Table 2.1
Member redundancy hierarchy for indeterminate structures given by Lloyd and
Member
Redundancy Member Classification
Level
p
-
- - -
A member whose failure leads to collapse for dead weight load conditions.
A member whose failure leads to progressive collapse for dead plus some fraction of live
weight load conditions.
A member whose failure leads to progressive collapse for a limited set of load conditions
that include dead and live loads in combination with some fraction of the design
environmental load.
A member whose failure leads to progressive collapse for a limited set of load conditions
that include dead and live loads in combination with some multiple of the design
environmental load.
A member whose failure has effect on the design strength, but whose presence
enhances the redundancy of nearby members, ie. a normally lightly loaded member that
provides an alternative load path when a nearby member fails.
A member whose failure has no bearing on the design, reserve or residual strength, ie. a
member.
In 1979 Marshal1 proposed two alternative measures of redundancy. For simple systems
with a number of identical parallel load carrying elements a redundancy factor (RF)
was defined as:
Values of RF less than unity therefore imply a high likelihood that initial failure will lead
to collapse, whereas very high values relate to damage tolerant structures.
The alternative measure is the damaged strength rating (DSR) given by:
damaged strength -
Eqn 2.7
intact strength
For more complex structures where is not directly available, the effect of damage is
established by comparing the results of structural analyses for the intact and damaged
structures. This will be illustrated in examples presented in Section 5.
In later work by et al (1988) another definition of redundancy, again denoted RF,
is defined as:
Ultimate structure resistance
RF = Eqn 2.8
Structure resistance at which first member fails
In some ways this definition may be considered to be more akin to the foregoing definitions
of reserve strength. Nevertheless, the notation indicates the strong correlation between
redundancy and system reserve. Use of this measure of redundancy, RF, will be
demonstrated in the presentation of specific results which follows. Caution in determining
first member failures is also required, however. It might be considered as the first
occurrence of plasticity which needs to be defined in terms of complete section or extreme
fibre conditions, or otherwise might be linked to buckling and a loss of a component
capacity. Recent use of the term 'System Redundancy Factor' (SRF) has
referred to first major member failure, to avoid reference to early failure of a secondary
component which plays no part in the overall system response. It should be noted that in
some instances first component failure may not necessarily be related to a member and
tubular joints and foundations may need to be considered.
In the jacket study by Nordal et al (1988) probabilistic measures were introduced and
additional consideration of these is given in Section 2.6. Taking and respectively,
as the safety index for the full system and for the union of first member failures (ie. the
combined probability of any member failing first), a redundancy measure:
2.9 CONCLUSION
In this section some key considerations in evaluating the reserve strength of jacket
structures have been introduced. Their simple definition has been shown to be complicated
by many factors arising from differences between design processes, structural configuration
and the purpose of assessments.
In order to provide some comparison in this review, the following terminology is adopted
wherever possible.
Reserve strength ratio
ultimate load at collapse
RSR = Eqn 2.12
design load
Redundancy factor
ultimate capacity
RF = Eqn 2.13
capacity at first member failure
Residual resistance factor
damaged structure capacity Eqn 2.14
RIF =
ultimate capacity
Responses are described as brittle when the overall load deflection curve for the structure
exhibits a rapid reduction, whereas ductility describes a gradual change in the resistance
curve.
In many instances sufficient information is not provided and alternative presentations are
necessarily adopted. Sometimes the obscurity is intentional as the results relate to
investigations for critical platforms. However, in all cases there is an attempt to explain
the work and its basis so that final comparisons can be made.
Having raised several areas of uncertainty at this early stage in the review, the relation with
the current literature can be explored and the various points are discussed further in Section
6.
Y CAPACITY OF FRAME
Measured response
Figure 2.1
Definitions of reserve and residual strength
FIG.
FRAME CONFIGURATIONS NOT MEETINGGUIDELINES
CONFIGURATIONS GUIDELINES
2.2
API guidelines for ductile configurations
MAX.
CAPY.
IDEALLY DUCTILE
I TENSION MEMBER
YIELD
BRITTLE FAILURE
AXIAL STRAIN
Figure 2.3
Ductile and failure modes (taken from and Bea)
K-BRACE
X-BRACE
X-BRACE
TENSION
MEMBER
COMPRESSION
MEMBER
DISPLACEMENT
(NOT TO SCALE FOR
Figure 2.4
Series and parallel K- and X-bracing analyses and illustration of X-bracing reserve
(due to and
DUCTILITY RATIO
I
ID 20 3.0
(INCHES)
Figure 2.5
Ductility ratio quantifying plastic deformation capacity based on Gates et
Figure 2.6
Environmental load history for cyclic assessment due to and Tromans
Load Low cycle fatigue fracture
yield global
yield deformations
Deck displacement
b) collapse
Load
Initial
yield
displacement
c) Shakedown
Figure 2.7
Failure and survival modes under cyclic loads et
3. EXPERIMENTAL INVESTIGATIONS
A number of experimental programmes have provided physical data on the reserve and
residual strength of tubular frames representative of offshore jacket structures. Not all the
tests reported here were undertaken for this specific purpose, nevertheless relevant data (for
example from the first cycle in high stress seismic loading scenarios) have been extracted.
The available results are reviewed in turn in Section 3.1 below, and these are followed by
a comparison and discussion in Section 3.2.
In evaluating the reserve strength of test specimens, the allowable loading for a
to code requirements is calculated based on the specimen geometry.
The ratio of the ultimate capacity sustained to this design load then gives the reserve
strength ratio (RSR) for the structure. This approach is appropriate for simple test
structures but differs from the evaluation of offshore jackets. These are designed for a
range of loading scenarios and components are sized to give elastic utilisations less than
unity in all cases.
Furthermore, test investigations generally identify exact member properties in terms of
diameter, wall thickness and material yield stress based on tensile coupon test results. In
the assessment of jacket structures only nominal or minimum specified properties are
generally available. However, to adopt these in the assessment of test specimens, although
attractive for giving a comparative basis, may not be practical. For example, test
specimens are loaded slowly and the results from static tensile coupon tests are more
relevant than from conventional dynamic testing. Material grades at the scale of testing
may differ and the associated distribution of yield stresses may not be the same. Similarly
ERW pipe is often adopted but is annealed after delivery, the minimum specification would
therefore not be an appropriate reference.
Caution should therefore be exercised when comparing ultimate strength evaluations for test
structures (based on measured properties) with predictions for jacket structures (using
nominal values) to ensure that this source of reserve strength is properly accounted for.
It can be seen that the stockier members in the shallower B series trusses ensure that the
structural response is more ductile. There is additional reserve beyond first member failure
before the peak load is attained (B-2 versus A-2).
Grenda, and Shinners
Static pushover tests were carried out at the Southwest Research Institute on six planar, K-
braced, single-bay tubular frames, some 9m by in size, and with gravity acting out of
the plane of the frames. The configuration of the one-third scale test specimens of Bass
Strait platforms, is shown in Figure 3.4. Lateral load was applied at the top of the frame
and reacted at the base by pinned supports. In test specimens 2 and 3 the can thickness at
the K joint was 0.432 inches whereas in tests 1 and 4 the can was omitted, giving a chord
wall thickness of 0.156 inches, equivalent to the member geometry. The braces were
grouted in specimens 5 and 6. Table 3.3 summarises the test matrix for specimens 1 to 4.
Table 3.3
K-braced frame configurations and capacities
4 Can 175
In all four tests the ultimate frame load (given in Table was governed by compression
member buckling. The overall load deflection responses of the frames are given in Figure
3.4. It can be seen that as the compression brace diagonal subsequently sheds load, the
lateral load resistance of the frame declined with further frame displacement, with a
consistent post-failure stiffness from one test to another. The third frame was deformed
far enough to mobilise the portal strength of the legs which gave a residual strength some
two thirds of the peak value.
The insensitivity of the responses to the thickness of the K joint cans is attributed to the
degree of overlap; much of the compressive brace load was transferred directly into the
overlapped tension brace, thereby reducing the stresses in the can and thus its influence on
K brace strength. The response has in fact spawned further investigation of the relation
between frame behaviour and isolated joint capacities (Connelly and Zettlemoyer, 1989).
This latter numerical study is described in Section 5.1.
The effective lengths of the compression diagonals, which governed the buckling load and
thus the overall capacity of the test frames, were calculated from the measured curvatures
of the member. Details of the compression member responses are given in Table 3.4.
Table 3.4
K bracing behaviour in test frames
BOMEL
Within Phase of BOMEL's Frames Project, four single-bay K-braced frames (Figure 3.5)
in which the central K joints were the critical component were loaded monotonically to
collapse. Frames VII, and X contained gap K joints, whereas the central node in
Frame IX was a concentric overlap joint (Table 3.5). Load was applied in displacement
control at the top of the frames and the supports were pinned at the base. Frame is
shown prior to testing in Figure 3.6.
Table 3.5
Details of BOMEL's single bay K-braced test frames
= diameters = diameter =
= chord = brace-chord angle
The global responses for the K-braced frames are presented together in Figure 3.7. It can
be seen that the frames with gap K joints continue to sustain increasing load until such time
as cracking occurs in the gap region due to shearing action across the joint (Figure 3.8).
Load is then rapidly shed as the load path through the K bracing is destroyed. Residual
capacity is dependent entirely on the surrounding structure, which for the test frames
constitutes only the frame legs. Table 3.6 compares the API RP 2A design loads for the
critical joints calculated allowing a overstress, with the peak brace loads transmitted in
the tests. These values indicate the level of reserve implicit within the component design.
In Frame VII, for example, it can be seen that the maximum load sustained by the critical
joint was compared with a design load to API RP 2A of giving a 'component
RSR' of 2.89. On the basis of this 'reserve' implicit within the component design, it is
expected that the load sustained by the frame in the test would exceed the design value as
a result.
Table 3.6
Comparison of component and frame capacity reserves
For the bracing and vertical legs of the K-framed tests, simplified assessment of axial
loads equates the design frame load to the design load for the component. On that basis
the Frame design load to API RP 2A is also whereas in the test the peak frame
load was giving an RSR of 3.25, exceeding the component value. However, as
noted above, this significant reserve is due in part to the conservatism of the design criteria
for the critical joint. To obtain a measure of the system or frame contribution to the
ultimate frame response an alternative assessment is proposed. It is the frame RSR beyond
the component conservatism which reflects the frame contribution and in percentage terms
this is given by
Frame RSR - Component 'RSR'
Frame RSR
For Frame the percentage frame contribution is therefore (3.25 - 2.89) 10013.25 =
11.1 % as shown in the final column of Table 3.6.
This presentation illustrates the difficulty in relying solely on the RSR to quantify the ability
of structural configurations to sustain loads in excess of the design value. The single bay
K frames in the BOMEL tests exhibit a high RSR but this is derived largely from the
conservatism in the design of the critical component and little contribution due to nonlinear
frame action is mobilised. An alternative frame configuration may offer alternative
loadpaths for redistribution contributing to an equivalent frame RSR but with lesser
conservatism in component design. The percentage is therefore a means to demonstrate the
contribution from nonlinear frame action to the ultimate system response.
There are significant differences in some joint capacity equations in API RP 2A (1993) and
the HSE Guidance Notes (1990). These arise due to factors such as different source
databases, the chosen lower bound or characteristic philosophies and definitions of first
crack or ultimate load failures. Depending which provisions are adopted, the apparent
RSRs vary. For Frame X, for example, the API joint and frame RSRs of 3.63 and 4.28
would become 2.06 and 2.43 if HSE Guidance were the base. This difference is
attributable to the accuracy of tubular joint equations rather than the inherent reserve
strength offered by the frame. By focusing on the reserve beyond the component
contribution the API ((4.28 - = 15.2%) and HSE (2.43 - 2.06) 10012.43
= 15.2%) based assessments reveal a consistent level of system reserve due to frame
action. In this way the percentage frame contribution may be considered to give a robust
measure of system reserve.
In Figure 3.8 the brace loads for Frame are compared with the global frame load. It
is clear that in the elastic region the components are equal, but the reserve capacity of the
portal action is utilised in the regime to enhance the overall capacity.
By contrast, the overlapping joint in Frame IX failed in an unexpected mode, with local
buckling of the brace wall at the compression intersection (Figure 3.10). The joint
remained intact and therefore imparted much greater post-peak residual capacity than the
gap K frames, as shown in Figure 3.7.
It may be concluded that although component reserves or inherent conservatism are large,
the contribution of frame action is not very significant. Furthermore, the tests confirm the
importance of a proper understanding of tubular joint behaviour within the confines of a
frame for structural collapse predictions.
Popov et
The results of two two-bay X-braced frame tests undertaken at the University of California
Berkeley in the late 1970s are reported and discussed in a number of references (Mahin et
1980; Popov et 1980; Zayas et al, 1982; Popov et al, 1985). Two tests (Frames
I and were undertaken as shown in Figure 3.11 with a prescribed sequence of cyclic
loads. Because the loads are reversed and incremented to collapse, direct comparison
cannot be made between these tests and other ultimate strength tests reported in this section
because of the uncertain influence of the development of plasticity on the peak load.
Nevertheless the influence of component slenderness on the response characteristics is
instructive. Table 3.7 presents the yield stress values based on tensile coupon tests and
section sizes from which the relative slenderness of Frame I compared with Frame is
evident. In the tests this was manifested in earlier and more severe local buckling and
tearing failures in Frame I whereas Frame was able to maintain its capacity for a greater
number of cycles and attain a larger lateral displacement thereby absorbing more energy
as required under seismic loading.
Table 3.7
Tubulars used in frames tested under loads at University of Berkeley
Frame I Frame
Member Yield
D (mm) T (mm) D T (mm)
The tests demonstrate the significance of relative section sizes and not just bracing
configuration on the inelastic response of structures. Because of the manner of load
application the tests are not considered further in this review of static reserve strength,
neverthess note of the frames is made given the scale of testing and the wide recognition
of the tests in the offshore industry. The work demonstrated the value of large scale frame
tests and led into the SCI (1990) and BOMEL (1992) programmes to investigate ultimate
static responses.
SCI
The four two-bay X-braced frames shown in Figure 3.12 were tested to collapse with
lateral load applied to the top of the frame under displacement control (SCI, 1990;
Billington et al, 1991 and Bolt et The bases of the frames were
hinged. Frames I and were designed for member failure. The horizontal brace was
omitted in the latter case to investigate the influence of the member which is lowly utilised
in elastic analysis on the collapse mechanism. Frame was specifically designed for the
compression X joint in the top bay to be critical. Frame IV included a fatigue crack to
demonstrate the influence on the collapse mechanism. Table 3.8 summarises the frame
configurations.
The primary bracing members were from BS 3601 ERW tubulars which were annealed to
remove residual stresses which had influenced the response of the K-braced frames tested
by Grenda et al. Stub column tests confirmed that the annealed tubulars gave a distinct
yield typical of offshore grade steels and that static coupon tests gave yield values
representative of member responses. The yield stress levels for the bracing averaged
287N/mm2, for the API 5L X52 legs 359N/mm2, and the thickened joint cans 324N/mm2.
The individual test results are described on the following pages.
Table 3.8
Details of SCI two-bay X-braced test frames
Figure 3.13 shows Frame I at the start of the test. Frame I was designed with a thickened
joint can at the top bay X joint (Figure 3.12) so that the compression diagonal was the
critical component.
The global response during the test is shown in Figure 3.14. This figure presents the
overall displacement of the frame as measured by a transducer at the top of the frame
parallel to the application of load. Scan numbers are marked on each plot and will be used
in reference to discussion of the test. A similar format is adopted in the description of
subsequent frames. The design load for the frames, based on elastic assumptions and the
capacity of the critical component given by API RP 2A provisions, are also shown in the
figures. Extreme 'storm' loading is considered as the reference for reserve strength
calculations with a one third increase in the allowable capacities accounted for.
The response of Frame I remained linear up to Scan 7. Subsequently the stiffness of the
frame reduced with further application of load, as the stiffness of the tension brace reduced
with the onset of tensile yielding. This can be seen from the forces recorded by the
integral brace load cells (Figure 3.12) plotted in Figure 3.14. This yielding before buckling
of slender braces with the same nominal properties, indicates the presence of a locked-in
pre-tension from fabrication (see also Briggs and 1978).
The lower portion of the top bay compression braces buckled at Scan 11 (Figure 3.15) with
a drop in the sustainable load (Figure 3.14). The buckle was at approximately to the
in-plane and out-of-plane directions. Two sudden drops in the load with increasing
displacement coincided with local brace wall buckling and crimping at the plastic hinge
locations at the leg tubular joint and at mid span of the buckled member.
Although the load in the tension member remained constant at yield and the load in the
damaged compression member reduced, the residual strength of the frame gradually
increased from Scans 13 to 17. Shear forces in the legs increased as portal action
developed and the axial load in the horizontal member increased. This illustrates the
importance of this member to the residual strength of the structure in the event of damage
or overload.
The comparisons between the design 'storm' loads that it is calculated the component and
frame can sustain in accordance with API RP 2A and the experimental capacities are
presented in Table 3.9. For Frames I and in which brace buckling dominates, the
for each would be the same whether API RP 2A or HSE Guidance Notes guidelines were
adopted. However, in cases where joints are the critical component (eg. Frames and IV)
the base design loads (capacities) are different giving apparent differences in the reserves.
In fact, for the 1.0 X joints in Frames and IV the API and HSE provisions give
almost identical capacities. However, for the general situation it should be noted that the
RSR calculated depends on the validity of the base formulation for predicting component
capacities.
Table 3.9
Comparison of component and frame capacity reserves from SCI tests
The conservatism in the individual component designs (expressed here as the component
RSR) would be expected to result in a corresponding increase in the frame capacity. Any
increase in the RSR beyond the component value is therefore attributable to nonlinear frame
action, alternative loadpaths etc., arising from the redundancy of the frame configuration.
Expressed as a percentage of the frame RSR this gives the percentage frame contribution
in the final column of Table 3.9. Further discussion of this approach was given in
conjunction with the BOMEL K-braced frame test results.
The results for Frame I are unexpected as the component appears to have greater reserve
than the frame. However, the member properties and critical loads for brace buckling and
tensile yield are similar. The addition of locked-in tensile loads from fabrication
determines that tensile yield occurs first and gives an increase in the apparent load at which
compression buckling occurs. In this case it is not therefore appropriate to evaluate the
frame contribution with respect to this 'apparent' component capacity.
This may not be representative of offshore structures where the welds, degree of constraint
and therefore shrinkage differ from scaled test structures. However, understanding this
influence, the contribution of X bracing action can be revealed. First yield occurs between
Scans 7 and 8, nevertheless the frame sustains increasing load, albeit at a reduced stiffness
as load is carried by the alternative compression diagonal. It is only with the second
component failure (brace buckling), that the frame capacity is exceeded. The ratio of peak
load to the frame load at first failure Scan 8 is 1.10, giving a measure
of the redundancy factor et al, 1988) due to the frame action.
Frame
Frame was the first frame in which a joint was the critical component. The top bay X
joint can was of similar thickness to the brace but had a higher yield stress. In all other
respects Frame was nominally identical to Frame I as shown in Figure 3.12. The overall
response of the frame is shown in Figure 3.16. The top bay compression X joint response
softened from Scan 6 and chord wall ovalisation became visible. However, this is not
apparent from the global load deflection response (Figure 3.16). because the tension load
paths compensated for the softening compression joints, carrying a greater proportion of
load. The limiting load was reached across the joint at Scan 11. The apparent capacity
was some 40% higher than had been predicted by reference to the mean of test data (HSE,
1990). This unexpected behaviour was due to the membrane action of the frame including
symmetry, and locked-in tension in the compression braces from fabrication.
The frame test was continued beyond first joint failure and yielding of the tension chord
spread from Scan 12 giving rise to a plateau in the response curve, with local yielding of
the bottom bay compression member noted at Scan 13. From Scan 16 onwards, yielding
of the tension brace became extensive, portal action developed in the legs and the brace
welds came into contact across the flattening X joint (Figure 3.17) allowing the joint to
transmit greater loads.
Driven by in-plane bending arising from the leg moments associated with portal action, the
top bay compression brace buckled predominantly in-plane at Scan 22 and the frame
capacity reduced. Figure 3.18 shows Frame was extremely ductile and the process of
load redistribution led to a higher capacity than for Frame I, albeit achieved at larger
overall displacement, as shown in Figure 3.19. This was not an obvious result as Frame
with a thinner X joint chord, contained less steel than Frame I. This test demonstrates
the potential importance of nonlinear joint behaviour on ultimate structural responses.
The role of frame action on reserve strength can be seen from Figure 3.16. The joint
capacity is reached at Scan 11 yet the frame load continued to increase until the capacity
of the alternative tensile load path through the X bracing was also reached. Both
compression joint failure and tensile yield are ductile modes of failure without rapid
reductions in capacity. The global frame response is therefore for the capacity to increase
as portal action is mobilised. Additionally the deformations across the joint enable a new
stiff load path to be developed as the braces contact, until buckling and rapid unloading is
precipitated. This gross nonlinear response is not predicted in elastic analysis. Comparing
Frames I and (Figure 3.19) it can be seen that the joint failure protects the compression
brace from early buckling.
The deformations enable parallel load paths to be mobilised such that when brace buckling
occurs portal action in the legs develops whilst loads are transmitted simultaneously by the
braces. The reserve strength beyond the critical component load is therefore considerable
as demonstrated by the large percentage frame contribution in the final column of Table
3.9. This contrasts with the figures for Frame below and the less redundant K-braced
frames (BOMEL) in which first member failure triggers frame collapse. It should be noted
that locked-in stresses have little influence on the global reserve. Tensile prestress exists
in both braces reducing the apparent tensile capacity and increasing the capacity in the case
of compression. The net effect on the global reserve is therefore negligible.
The important role of joint nonlinearity on frame behaviour is clearly revealed by this test.
Frame
Frame was identical to Frame I, except that the mid-height horizontal was omitted
(Figure 3.20). The member carries negligible load in the elastic regime (Figure 3.14) and
might be omitted in practice to reduce structural weight. The implications of reduced
redundancy on reserve and residual strength were therefore examined in Frame
The global response is shown in Figure 3.21. The response was linear up to Scan 5. With
the next increment the top bay lower compression brace buckled, shedding load via the
alternative top bay (tension) diagonal directly into the bottom bay compression member,
which became visibly bowed. The alternative load path ensured that the overall frame
capacity was maintained with no rapid reduction in load. From Scan 8 the upper bay
tension member began to yield and the bottom bay compression member buckled with
increasing displacement at Scan 1 1 . Figure 3.22, looking down the top bay compression
brace, shows the dual member failures as well as the portal action mobilised in the frame
legs.
The sequence of failures, without the mid-height horizontal to evenly distribute loads to the
bottom bay, reduced the residual frame capacity below the design storm load. Figure 3.23
compares the Frame I and Frame results. Larger imperfections and compressive
in stresses, relating to the degree of restraint and specific fabrication procedure adopted,
also explain the difference in peak capacity. However, the role of the mid-height horizontal
in maintaining the residual strength is shown and this is also indicated in Table 3.9 by the
difference between the frame and component reserve strengths.
Frame IV
Frame IV was nominally identical to Frame except that a fatigue crack had been
introduced at the critical top bay X joint (Figure 3.24). The global load-deflection response
of Frame IV is shown in Figure 3.25. At Scan 8 ovalisation of the X joint chord became
visible, there being relative displacement between the fatigue crack faces as the brace-side
crack face slid under the chord side. The X joint response is shown in Figure 3.25 from
which the capacity can be seen almost equal the uncracked joint capacity in Frame
(Figure 3.16). By Scan 1 1 the upper portion of the compression brace was pushing into
the chord under the existing crack and by Scan 13 the two braces were in contact (Figure
3.26).
As the crack propagated round the weld toe, the chord began yielding due to the reduction
in its cross sectional area. At Scan 20 a tear opened perpendicular to the axis of the chord
and rapidly propagated, rupturing the chord and the global load fell. This changed the
compliance, and with additional prescribed displacement the compression brace lifted at the
joint.
The crack had very little effect on the global response as seen in the comparison between
the Frame and IV responses in Figure 3.27 and Table 3.9. However, the reserve
strength associated with joint failure, even with the presence of a crack, is confirmed. Had
the crack been oriented differently the effect may have been for fracture to occur earlier
with a rapid reduction in capacity and with less action having been mobilised to
contribute to residual capacity.
BOMEL
In Phase of the Frames Project two additional double bay X-braced frame tests were
undertaken to further examine the influence of joint behaviour on system responses. The
frames are shown in Figure 3.28 and Table 3.10 summarises their purpose.
Table 3.10
Details of BOMEL two bay X-braced test frames
Frame V was not tested to collapse. It confirmed the influences of locked-in stress as
postulated for the SCI tests above. The test was halted when the applied load gave a
multiple of 3.1 on the design load. The global response was ductile again indicating that
the RSR would considerably exceed 3.1.
Frame
Frame was fabricated by replacing the failed joint in Frame V with a new joint of
reversed configuration (Figure 3.28). Figure 3.29 presents the global load-deflection
response and the member responses, showing the distribution of loading between the tension
and compression load paths within the top bay X-braced panel, are shown in Figure 3.30.
The response of the frame and joint were ductile, contrary to expectations for a tension
loaded X joint, in which crack initiation and fracture usually occur. These expectations are
derived from isolated tests where brace loads but no chord loads are generally applied.
However, a high level of chord compression was developed in the Frame joint as the
tension path through the joint softened, and this contributed to the yielding and gross plastic
deformation at the joint. At the end of the test, although grossly distorted, the steel joint
remained intact.
The member forces in Figure 3.30 illustrate clearly the contribution of X-braced frame
action on redundancy in comparison with the K-braced frame responses in Figure 3.7. As
the tension joint softened initially, so additional load was carried by the alternative
compression diagonal, giving an imperceptible change to the stiffness of the global
response. However, as buckling of the compression chord took place, so the increased
deformation enabled greater loads to be transmitted in membrane action across the 1.0
tension X joint. Only when significant portal action, full brace yield and plastic buckling
of the compression braces were acting, was ultimate frame capacity attained. The frame
is shown at this stage in Figure 3.31. It should be noted that the X joint did not crack in
tension, possibly due to the presence of high chord loads, but deformed in a ductile manner
(as shown in Figure 3.32 where cracking of the paint is observed). The reserve strength
and comparison with frame design loads are given in Table 3.1 1.
1
Comparison of component and frame capacity reserves
Comparisons both for API RP 2A and the UK HSE Guidance Notes are presented,
illustrating the dependence of the RSR on the base design criteria. There are significant
differences between API and for tension joints and these are manifested in RSRs of
6.58 compared with 3.72. However, isolating frame action by comparing the global and
component RSRs in both cases, reveals a consistent contribution of some 20%.
This strong influence of frame action ensuring system reserve is a function of the frame
being X-braced and first failure being joint related.
lnoue et
The effect of a horizontal brace on the performance of X-braced tubular frames was studied
in this series of ultimate load tests of two plane frames and two space truss specimens. The
structural configurations, together with the member dimensions and properties, are shown
in Figure 3.34. The members were manufactured from ERW pipe. The joints were not
tube-to-tube welded connections as used offshore; instead brace members were welded to
gusset plates attached to the chord. This method of attachment gives fixed end conditions
in the plane and pin-ended conditions out of the plane of the gusset. Nevertheless the
comparative tests with and without horizontals and in two and three dimensions are
instructive. The specimens were cycled beyond the attainment of a peak capacity but the
first half cycles are of relevance here.
The two dimensional frames were loaded in-plane, whereas the space frame was loaded
across the diagonal. Against each response curve, in Figure 3.34, the legend enables the
progress of failure to be followed.
Considering first the planar frame behaviour, it can be seen that buckling of the
compression braces initiated collapse but it was only with failure along the alternative load
path that the peak load was attained. The maximum load sustained by the frame with
horizontals was some 10% greater than that by the frame without horizontals.
Furthermore, the drop in load was more abrupt and resulted in a much smaller residual
capacity in the latter case. The different behaviour is attributable to the horizontal brace
which took significant load from the tension brace as redistribution from the buckled
compression brace occurred. Where the horizontal is omitted a second buckle in the lower
bay was precipitated. This correlates with the findings for Frames I and in the SCI tests
(SCI, 1990; Bolt et al, 1994).
The inelastic behaviour of the space frames subjected to direction load was dominated
largely by buckling and yielding of the legs, rather than the braces as in the plane frame
tests. Accordingly the horizontal members made a less significant contribution to the
overall load-deflection response. Nevertheless in the absence of horizontals the peak load
was about 10% less than in the fully braced case. Direct comparison with the 2D tests is
not possible because of the re-orientation.
Tests (Soreide et all
Two X-braced three dimensional frames were tested by SINTEF to ultimate load as shown
in Figure 3.35 a and b. Load was applied along the axis of member EF whose wall
thickness was increased accordingly to obviate buckling. The frames, denoted S1 and S2,
had different ratios and brace slendernesses and in both cases the joint cans were
thickened to ensure member failure dominated.
Soreide et al (1986) presented USFOS (analytical) predictions for the frame responses in
advance of the test work being carried out. These are shown in Figure 5.23 indicating
(factors on the design load) of around and 3.3 and 3.6 with a considerable margin
between first component failure and the peak frame capacity. The basis of the 'design
loads' indicated by the authors is not given and so absolute comparison with other test
programmes must be made with caution.
The actual test results reported by Soreide et al (1987) are also given by the dashed line in
Figure 3.35 in terms of the global load applied and the lateral displacement of the frame
at E. From the graphs in Figure for Frame it can be seen that the stiffness
degraded gradually prior to failure, when buckling occurred in member EB at an applied
load of about At that point member IF, which was at yield, fractured, due to lack
of fusion in the weld to the can at Joint I, causing a sudden drop in the external load. With
redistribution some additional load was sustained but a similar fracture then occurred in
member JG at joint J precipitating another abrupt decrease in load bearing capacity. At this
point the test was terminated. Fracture did not occur throughout the second frame test, S2.
The diagonal bracing in Frame S2 was more slender than it had been in S1 and buckling
occurred first in member EI. A small local dent (about 2mm deep) had been introduced
in this member during transportation which, given its ratio of 63, was vulnerable to
imperfections. Subsequently the frame continued to take load but with an ever reducing
stiffness.
Frame S1 demonstrated the 'brittle' response related to fracture where the rapid loss of
stiffness along the load path cannot be compensated by redundancy within the simple
structure. In comparison with the original Soreide et al (1986) prediction in Figure 5.23
it appears that the frame sustained a load around (slightly higher than predicted for
the perfect structure), giving an apparent RSR of 3.3. However, it is likely that the
Soreide predictions were based on nominal material properties and dimensions whereas
other tests have been interpreted on the basis of actual properties. Furthermore the basis
of the 'design load' is not known. The deformation characteristics of the frame with initial
nonlinearity occurring around and a peak load corresponding to a deflection in the
range are similar in both initial analytical prediction (Figure 5.23) and test.
However the later analysis and test comparisons in Figure offer less good agreement,
although the stages of rapid unloading associated with fracture have been tracked.
In Frame S2 however, the global response maintained a positive slope as alternative bracing
load paths compensated for the falling load. The original Soreide et al (1986) prediction
indicated a peak load of achieved at around 40mm deflection, whereas in the
presence of the dent the ultimate load was associated with a deflection of some
It may therefore be concluded that the RSR was as predicted, ie. around 3.6, but the
reservations noted for Frame S1 apply together with concerns about the acceptable
deformations to be associated with the RSR. Figure shows the comparison between
the 'dented' prediction and experiment. From the figures the responses at component level
provide no explanation for the apparently higher stiffness determined analytically, compared
with the gradual softening in the test.
Comparative analysis
Many of the frame tests described in this section have been used as a baseline for
calibrating nonlinear analysis programs. Discussion of these results would provide little
additional information on the reserve strength characteristics of tubular frames, however
Table 3.12 identifies references where test and analysis comparisons are presented. These
are clearly important stages in the validation of software for predicting the ultimate
responses of jacket structures (see also Section 4).
3.2 COMPARISON OF RESULTS
The individual experimental results described in Section 3.1 are summarised together in
Table 3.13. Measures of reserve strength are presented. The form depends on the
information given in particular references. Comparisons are made in the sub-sections
which follow.
K versus X bracing
Comparison of the Grenda and BOMEL K frame tests with the X-braced frames,
demonstrates that in the first case only the surrounding frame (ie. legs) can contribute to
the reserve strength. In the second case the bracing within the panel presents an additional
source of reserve and it is only when the limiting capacities of both brace load paths have
been reached, that reliance is placed on the legs.
The magnitude of the X-braced panel contribution, depends on the relative slenderness and
therefore the capacity of the braces with respect to yield and buckling. Similarly the
relative contribution from the legs depends on their stiffness. This is influenced by the
length of the legs and therefore the contribution from portal action would be different from
the BOMEL K frames and SCI-BOMEL X frames despite the leg tubulars being of the
same size.
On this basis direct comparison of for X and K frames is not meaningful. The
important point demonstrated is that the X framing offers an additional source of reserve
strength over and above K action. This is reflected in the recommended bracing for
framing of structures in earthquake regions embodied in API RP 2A (Figure 2.2).
Table 3.13 Summ of experimental investigations of m reserve strength (Part 1 of 3)
FAILURE MODE REFERENCE LOAD-DEFLECTION SYSTEM RESERVE NOTES
RSR %frame action
Briggs Maison Relatively large legs
Significant portal action
Ogawa et
et Buckling insensitive to
changes in chord at K joint
BOMEL
3.13 Summ of experimental investigations of system reserve (Part 2 of 31
SCI
BOMEL* V 3.1
6.58 API
3.72
Aside from the experimental uncertainties noted, a consistent view of measured effective
length factors being some 30%less than values given in RP 2A is shown in Table 3.14.
Relating these experimental findings to those analysis methods (eg. INTRA) for which
effective length factors have to be specified by the analyst (see Section the importance
of good data for reliable structural collapse load predictions is clear. A dependence on
code values would not give reliable reserve strength predictions.
3.2.9 Materials
In configuring test frames to represent offshore jacket structures a principal constraint is
the type, grades and produced sizes of steel, given the limitation in terms of frame size and
capacity that can be accommodated for a reasonable budget. An additional factor is the
stress-strain characteristics of the chosen steel in relation to 'typical' offshore materials and
the influence of girth welds, longitudinal welds, etc., on component capacities.
The materials used in the test frames have been reviewed and are summarised in Table
3.15. In general the references give insufficient detail for conclusions about member
properties to be drawn. The exception is the interaction between the Grenda, SCI and
BOMEL tests.
3.15
Comparison of steels in test frames
From the Grenda tests it was recognised that the fabrication of tubulars in the electric
resistance welding (ERW) process introduces residual stresses which cause a larger portion
of a member cross-section to yield at a given applied load, thus reducing capacity. Having
quantified this effect in the Grenda tests, steps were taken to eliminate these residual
stresses in the SCI and BOMEL tests by annealing the tubulars prior to frame fabrication.
Stub column trials showed the annealing process to restore the onset of yield (nonlinearity)
from about 50% of the plateau stress in the as-welded condition to around
engendering the yield characteristic of offshore structural steels. The series of stub
column tests also showed that girth welds (present in full scale jackets) had no apparent
effect on the response (Bolt et al, 1994).
The results presented by Grenda et al include a correction for the ERW
properties. However, without further evidence for the treatment state of the Briggs and
Maison tubulars it remains a possibility that the early compression failure and ductility were
influenced by residual stresses from the ERW process.
Changes to the characteristic of the response of components and hence test frames, need
to be allowed for in the relation of the results to jacket structures. Furthermore in
performing calibration analyses, appropriate stress-strain characteristics need to be
modelled. This section therefore provides a caution to the immediate adoption of test
results unless full details of the geometry, material and fabrication are known.
As an alternative to material specifications some authors report specific tensile coupon
results. The rate of testing can influence the yield value recorded and therefore the coupon
testing procedure needs to be linked to the rate of load applicationto the structure. Practice
is not uniform across the test programmes, however.
0 Conclusion
The general findings from experimental work will be drawn together with numerical results
in the discussion in Section 6.
i
TO GRADE
STRENGTHS ARE BASED ON DYNAMIC
TENSILE COUPON TEST RESULTS
LOAD CELL
PRESSURE
J
0 1.0 2.0 2.5 3.0
FRAME DEFLECTION
Figure 3.1
and Maison single bay X-braced test frame
Brace Nos
figure 3.2
trusses tested by Ogawa et
Figure 3.3
Truss load deflection response (Ogawa et
Figure 3.4
Single bay K-braced frames tested by Grenda et
FRAME FRAME
FRAME FRAME X
Figure 3.5
Single bay K-braced frames tested by BOMEL
0 20 P
Figure 3.7
Global load-deflection responses for BOMEL K-braced frames
FRAME
Figure 3.1 1
Two bay X-braced frames tested under cyclic loading by Popov et
FRAME DISPLACEMENT (mm)
Figure 3.14
SCI Frame I global and local member responses
LOAD
X
W
0
m
W
F
I V MODIFIED
FRAME V FRAME
Figure 3.28
Two bay X-braced frames tested by BOMEL
P-
Necking at joint.
brace bowed.
50 100
Frame Displacement (mm)
Figure 3.29
BOMEL Frame global response
TENSION BRACES
COMPRESSIONCHORD
Figure 3.30
Frame member forces showing X bracing load distribution and frame action
C
-: Present Experiment
Present
, l
,
0.0 10 30 40 SO 60 70
point displacement
pin .
,: Present Theory
0.0 20 70 90
displacement
Figure 3.33
Frames tested with and without damage by Paik and Shin
Oil
Reaction
PC
Oil
Cell
Figure 3.34
X-braced frames and towers tested by lnoue et
-- C cc-
--*
Figure
3D X-braced box frame tested by Frame S1
failure due to buckling and fracture due to lack of fusion
- o
LOAD
LOAD
Figure
3D X-braced box frame tested by Frame S2 -
buckling failure initiated by dent in member
4. SOFTWARE FOR
PUSHOVER ANALYSIS
-
4 . 1 . 3 Linear superposition strain based
As an advance on the member replacement method Holnicki-Szulc and Gierlinski (1989)
proposed a method which uses constraint equations based on lack-of-fit member strains,
providing a phenomenological capability. In combination with the linear solution,
the nonlinear structural response can be modelled. The concept uses the superposition
principle to combine the linear elastic solution with 'virtual' distortions introduced in the
structure, hence it is known as the Virtual Distortion Method(VDM). In addition to the
elastic stiffness matrix, a matrix representing the sensitivity of the structure to local damage
such as hinge formation or fracture is required. The matrix components are given by
deformations at a member end due to unit deformations at the other. The virtual state is
scaled such that global equilibrium and limit strength conditions are satisfied giving an
approximation to the condition within a progressively collapsing structure.
Failure is determined by the formation of a collapse mechanism when the permanent
deformations (virtual distortions) grow infinitely.
4.1.4 Linear superposition - load based
and van de Graaf (1990) developed the linear superposition method further and
their load based method enables additional constraining loadcases to be defined simply.
The method is restricted to axial nonlinear behaviour but can model cases of practical
importance where axial brace buckling or yielding, or axial pile failure dominate the
response. Instances of leg failure due to beam-column action cannot be simulated.
Figure 4.1 illustrates the method whereby the difference between the linear and nonlinear
responses of a component are identified. The difference is accounted for by applying end
force pairs to each nonlinear member. The constraint equations determine the correct
deformations and calculate the components of resistance to be derived from the member or
the surrounding frame. The nonlinear problem is reduced to finding a set of member end
forces, which, when combined with the applied loading, constrain the linear system to
follow the behaviour of the nonlinear system. The set of constraining forces can be quite
small as relatively few members (typically around ten) contribute to the collapse
mechanism. This approach is seen by and van de Graaf to have several advantages
over more complex nonlinear analysis programs. The authors consider that because a
linear structural model and a suitable linear analysis program are generally readily
available, a full collapse analysis of a complex structure can be performed within a few
days, in contrast to the months required to regenerate the model and perform the analysis
using a more complex nonlinear analysis capability.
The method is advocated for sensitivity analyses or to determine critical scenarios for
complete nonlinear analysis (van de Graaf and Tromans, 1991).
RSR =
[S,,
where guidance on the scoring factors given by Bea and Craig is reproduced in Tables 4.1
and 4.2.
Table 4.1
Level 1 RSR capacity scoring factor guidelines and Craig,
Guideline Score
Structure and foundation design and construction criteria (in relation to API RP 2A)
- 1959
- 1964
1965- 1975
1976- 1993
Table 4.2
Level 1 RSR storm and operational loadings scoring factor guidelines
and Craig, 1993)
Level 2 is based on the limit equilibrium technique due to Bea and (1993)
described in Section At Level 3 linear elastic analysis is used where WSD criteria
are modified to give best estimate ultimate capacities and for LRFD structures the partial
resistance factors are set to unity. Platform capacity is based on the inverse of the largest
allowable stress ratios for the primary members (ie. those critical to ultimate capacity).
This is subject to a robustness factor to reflect the influences of system redundancy,
ductility and conservatism in primary components in terms of post yield and load
redistribution effects. Level 4 encompasses a full nonlinear analysis either within the
context of deterministic or reliability based approaches.
The focus in this review is on the reserve strength of jacket structures and to that end full
nonlinear analyses offer the route to the more accurate predictions. Nevertheless the
foregoing methods and other such simplified approaches to evaluating approximate estimates
of RSR offer a rational and cost-effective methodology for assessing system reserves.
Furthermore this sequential analysis is the methodology being proposed for the assessment
of existing platforms in the draft API RP 2A Section 17.0 (1993). There the progression
is from screening, through design level to ultimate strength level analysis, as required, to
demonstrate system adequacy. It is only at the third level that inelastic, static pushover
analysis 'may' be required. However, if nonlinear analysis is necessary it is likely that the
structure is at greater risk and therefore particular confidence in the analytical techniques
is required.
FACTS Finite Element Analysis and for Complex Structural Software Development, USA
Three dimensional Systems
Information on the various programs for ultimate strength analysis listed in Table 4.3 is
given below. The descriptions of each program have been provided by the associated
development or support organisations to the same specification. The specification asked
for a 500 word summary of the program describing the status as at the end of May 1993,
with information to be provided covering the following points:
Beam element formulations
Spring elements
Method of solution
Load application
creation facilities
Offshore code processing facilities
facilities
Ongoing developmentslenhancements relevant to collapse analysis of jacket
structures
Date of original development, current status and contact details for further
information.
The organisations were also asked to confirm that the current report provides
comprehensive coverage of publications involving the use of the software. These are
included in the reference section and are discussed in Sections 4.3 and 5 which follow.
In addition to the software description, diagrams could be supplied and the material
provided for each program is reproduced below without modification.
Although not strictly a 'nonlinear analysis package', details of the VDM module within
RASOS are also included in this section as several North Sea operators are participating
in the development of the package. The information was supplied by WS Atkins to the
same specification as for the nonlinear programs presented above.
EDP
"The Extended Design Program, EDP, performs general nonlinear three dimensional
element analysis of structural systems including the effects of soil-structure interaction.
The program is suitable for static and dynamic analysis of steel and concrete structures such
as fixed and compliant offshore platforms.
Element formulations
The EDP element library consists of many formulations for modelling the
nonlinear response of offshore platforms, including linear and
beams, struts, plates, shells, solids, pipes, quadrilaterals, loading and matrix
elements. Nonlinear elements are hysteretic and all elements are three dimensional.
Geometric stiffness, large displacement and P-delta effects are included.
Two and three hinge beam-column elements capture plastic hinge formation due to
combined axial force and bending moment interaction for primary bending members. The
strut element is a multipurpose phenomenological element to represent buckling and tension
yield in primary braces or large deformation cable tension response. Special composite
tubular beam-joint elements combine the nonlinear response features of the strut buckling
or inelastic beam-column elements with the joint nonlinear behaviour.
Plates and solid elements enable the finite element analysis of tubular joints, concrete shells,
interaction and diaphragm response.
Spring elements
Spring algorithms enable any physical force deformation behaviour to be replicated.
pile interaction, structural joints or boundaries and their energy absorbing characteristics
are incorporated by use of nonlinear elastic or hysteretic spring elements. Soil axial and
lateral stiffness nonlinearity are modelled using a hysteretic soil element that includes
degradation and Spring elements also model impact, uplift and friction.
Method of solution
EDP contains features which minimise the potential for numerical instability in static and
dynamic analyses. Constant stiffness and Newton Raphson iteration schemes are used for
both static and dynamic analyses to maintain internal equilibrium with the externally applied
loadings. Nonlinear dynamic response is determined by direct integration of the equations
of motion. Linear response is obtained from response spectrum or time history analyses.
EDP is capable of incorporating local failure with advancing time. The analysis then
continues with this revised state and thus simulates progressive collapse situations to the
point of ultimate instability. When considering such alternate load paths for redundant
structures, the failure criteria for brittle and ductile members may be defined by force
limits, or deformation limits, cumulative deformation or hysteretic energy dissipation.
A demonstrated characteristic of the program is a very high efficiency in numerical
solutions.
Load
Static and dynamic loads may be superimposed to replicate the actual instantaneous and
complicated loading sequence experienced by the system. Static loads
are applied in a prescribed manner by the user or automatically by the program in
increments. Dynamic loads can be time varying forces (nodal or element), displacement
functions, wave loads or ground acceleration time histories. Multiple support excitation and
phased motions simulate the spatial variability of seismic motions for extended structures.
Code checking
Tubular steel structural members are checked to API RP 2A code requirements, even
during seismic time history analyses. AISC code checks are available for other sections.
facilities
EDP has an integrated comprehensive graphics capability to facilitate interpretation and
documentation of results.
Current status
First developed in 1980, and continually enhanced, EDP has been used extensively in
production analysis of complex offshore structures and has been checked against other
nonlinear programs. EDP versions exist for a range of computers from through
mainframes.
A reference including the use of EDP is by Piermattei et al (1990).
For further information contact Digital Structures 2855 Telegraph Avenue, Suite 300,
Berkeley CA 94705, USA, Tel: 510 549 1565, +
1 510 549 1600.
FACTS
"FACTS is an integrated library of programs for the linear and nonlinear element
Analysis of Systems under static and dynamic loads.
Element
The FACTS library includes both general purpose and application specific element types
and internal modelling features including:
3D linear: truss
beam
straight and curved pipe
3-9 node membrane element
3-9 node shell element
8-20 node solid
3D nonlinear: truss
cable
beam-column, lumped plasticity
beam-column, distributed plasticity
degrading stiffness beam-column, lumped plasticity
straight pipe
curved pipe with effects
8 node shell
8-20 node solid
gap friction
shear degrading link
free-field soil
near-field soil
2D nonlinear 4-8 node membrane element
U-bar restraint
SSD nonlinear buckling strut
buckling strut
Method of solution
Linear static solution
Multiple static, thermal and pressure load cases.
Specification of loads at substructure level.
Automatic recovery of all substructure displacement and stresses.
Linear dynamic solution
Eigensolution using an efficient and reliable shifted iteration scheme.
Component modes for linear substructures.
Response spectrum dynamic analysis.
Modal superposition time history dynamic analysis.
static solution
Ability to include linear as well as nonlinear elements and to specify that certain
substructures remain linear analysis).
Newton-Raphson Iteration: Both load and displacement iteration options are
available.
Event-to-Event Strategy: Automatic subdivision of the load step based upon events,
strut buckling etc., to stabilise the analysis with minimal user interaction.
dynamic solution
Ability to specify linear elements and substructures.
Newton-Raphson Iteration: Step-by-step analysis with constant time step with or
without equilibrium iteration.
Automatic Event-to-Event Strategy: Automatic modificationof the integration time
step based on accuracy considerations to minimise equilibrium balance requiring
minimal user interaction.
Solution control
Pause and restart from any state.
Organisation as a series of computational models linked by commands which are
input as program data. This provides unlimited flexibility in ordering the input
data and computational sequence, while still allowing well-established sequences
to be 'hard-wired' by macro commands.
Load ion
FACTS includes a hierarchial database management system which allows the FACTS
analysis programs to exist as stand alone modules. Access to the database is through an
easy-to-use interface language that allows users to develop and link their own applications.
Analysis of Structural and Mechanical Systems - GENLIB - For static and dynamic analysis
of structures with localised nonlinearities, including structures with nonlinear foundations
and general linear and nonlinear static and dynamic finite element stress analysis. Static
loads include: nodal applied forces, self-weight and temperature and pressure distributions.
Dynamic loads include: applied force functions and multiple support seismic excitation.
Inelastic buckling and postbuckling analysis of pipelines and stiffened shell structures made
of metallic or composite materials.
Hydrodynamic Analysis - - Static and dynamic wave-structure interaction.
Frequency domain and time history analysis of the hydrodynamic response and wave force
distributions for floating or fixed based platforms in regular or random seas.
Seismic Analysis - SEISLIB - Seismic strength and ductility analysis.
Ice-Structure Analysis - ARTICLIB - Dynamic ice-structure interaction.
creation
Tubular Joint Analysis - - Automatic modelling of complex tubular joints in
offshore platforms to capture the effects of joint flexibility and to estimate the stress
concentration factors.
Geometry definition - Includes point, line, grid and solid generation schemes. FACTS also
provides general kinematic constraints on nodal degrees of freedom.
Multi-level substructuring - Provides up to 25 levels of substructuring with up to
substructures in each level. FACTS includes automatic transfer of stiffness, loads,
recovery of displacement and recovery of stresses. Free word input procedure.
Offshore code
Fatigue Analysis of Structure - - A post-processor to to
predict fatigue life and long term statistics. Performs stress determination, member
checking and joint can design subject to API RP 2A guidelines.
facilities
Full interactive with options.
Interactive preview of results saved.
Initial geometry plotting.
Deformed geometry plotting.
Construction and plotting of result tables.
Graphical presentation of displacement and element results.
Current status
The FACTS package can be made available on a wide variety of computer systems
including IBM, CDC, CRAY, and Sun SPARC.
Early application of the FACTS software is given in a reference by Bouwkamp et
FACTS is available for lease or purchase. For further information contact SSD Inc, 1930
Shattuck Avenue, Berkeley, California 94704, USA, Tel 510 849 3458.
KARMA
"The Computer program. KARMA, developed by ISEC Inc., is a three-dimensional,
inelastic, nonlinear, static and dynamic response analysis program. It is based on finite
element formulations for structures subjected to environmental loads. A coupled foundation
structure, 3D nonlinear dynamic failure analyses of land-based and offshore structures can
be performed. KARMA, an United States oil industry standard program, is a result of 14
years of in-house and oil industry sponsored research and development.
The program's capabilities can be effectively used to evaluate ductility, failure modes or
structural integrity of a variety of onshore structures (bridges, overpasses, high-rise
buildings, transmission towers, etc) and offshore structures (jacket type, jack-up rigs,
deepwater compliant structures, floating systems, tension leg platforms, systems,
etc). Wherever possible, the program's methodology has been verified using model scale
and full scale experimental data.
Linear and nonlinear dynamic analyses are performed in the time domain. Alternatively,
linear dynamic analyses for earthquake loads may be performed in the frequency domain
(response spectrum analyses). Environmental loads may be wind, hurricane, wave,
earthquake, boat impact, or any generalised loading. Eigensolutions of large 3D complex
systems may be performed using a state of the art shifted eigensolver. The
program can handle material (steel, concrete, reinforced concrete, soil) as well as geometric
nonlinearities. KARMA is computationally very efficient.
KARMA has an automated redesign capability for offshore structures based on a minimum
cost algorithm. Also, comprehensive fatigue analyses can be performed including all
nonlinear effects.
A comprehensive post-processor exists in KARMA. It performs member checks using API
RP 19th Edition approach for tubulars. Joint checks are performed using
the AWS Alpha approach for joint classifications. Non-tubular member checks are
performed using the AISC recommendations. It also includes the interface to the 3D
general purpose visualisationand animation program, SWAMI. Thus results from KARMA
can be visualised and animated using SWAMI. Alternatively, KARMA models can be
interactively built with SWAMI'S 3D graphics capability.
Some of KARMA'S special features are listed:
1. Conventional fixed format input or key word oriented input structure.
2. Input data echo and extensive data checking capability.
3. Automated generation of inelastic properties for truss and beam element types.
4. Profile minimiser to optimise storage and computations.
5. Flexible boundary conditions handling through or multipoint constraints.
6. A generalised damping formulation to allow user specified damping ratios in as many
modes as desired.
7. Automated earthquake load profile generation for static pushover analysis.
8. Automated wave load profile generation for pushover analysis.
9. Marine growth and comprehensive hydrodynamic options.
10. Extensive element library for structure, soil, foundation and specialised applications.
11. Flexible and comprehensive load definition options including member loads, random
wave generation for unlimited durations and dynamic wind gust loads.
12. Automated self-tracking static and dynamic analysis capabilities.
13. State of the art fatigue analysis options which include all nonlinear effects.
14. Comprehensive post-processing and interfaces to 3D photoelastic graphics and
animation with lighting and shading (Program SWAMI).
Karma can be executed on workstations such as the Sun Micro Systems and
minicomputers like the VAX series, mainframes such as the IBM, parallel
processors, or the super-computers such as the CRAY."
A reference including the use of KARMA is by Gidwani and Renault (1990).
Further information can be obtained from ISEC Inc., 505 Montgomery Street, Suite 680,
San Francisco, CA 94111 USA, +
1 4 421 1692.
Introduction
PMB began its development of state-of-the-art nonlinear analysis software for the oil and
gas industry in 1978 with the introduction of the INTRA
Analysis) computer program. INTRA was originally developed through a joint industry
project to meet the challenge of analysing platforms in seismically active areas where
extensive nonlinearity is expected and confirmation of adequate ductility is essential for
good design.
INTRA was expanded for analysis of a wide range of structural types and thus became the
standard for nonlinear dynamic analysis in the oil and gas industry. Additional
enhancements included the development of advanced soil-pile modelling capabilities which,
in 1983, represented a capability that is still unique within the industry. was
created in 1987 when the basic architecture of the program was updated with improved
solution schemes and computational efficiency.
General
is a comprehensive nonlinear static and dynamic analysis tool which can be used
to analyse a variety of offshore structures, including fixed platforms, compliant towers,
stiff and flexible risers, buoys and cable systems. It has been used
extensively to perform regular and random wave analyses, pushover (collapse) analyses,
ship impact analyses, and toppling analyses. Its basic capabilities include:
A library of linear and nonlinear finite elements.
Regular waves, irregular 2D and 3D waves, current, ice and earthquake loading.
Static, eigenvalue, frequency domain and explicit time domain analysis.
API code checks and fatigue life calculations through a graphic post-processor with
automatic redesign capability.
Interactive, graphical model generation and post-processing with colour animation.
Element library
In addition to the conventional linear beam, plate and truss element, SeaStar has an
extensive library of nonlinear finite elements which explicitly model material nonlinearity
and large deformations including, for example:
Beam column
bending failure surfaces are either generated automatically using an
extension of Mohr plasticity theory or user-specified.
Substructured beam-column
Represents material and geometric nonlinearity including explicit modelling of local
damage.
Pile-soil interaction
Discrete elements to model nonlinear axial and lateral interaction for both static
and dynamic loading including rate of loading effects, cyclic degradation, hysteretic
behaviour and
Properties are either generated automatically based on API formulations, using
algorithms calibrated to large-scale field tests, or are user specified.
Speciality elements
- A phenomenological model of brace buckling, post-buckling and cyclic
response.
Cable element - Represents nonlinear interacting components of cable response
with a single element.
element - Models contact problems such as riser-seafloor.
interaction
Wave, current and earthauake loading
Wave loading in SeaStar includes regular wave, irregular waves, current loading,
kinematics stretching and wave-current interaction effects. Regular waves can be modelled
using Airy waves, Stokes fifth order waves, or Stream Function waves. Irregular waves
can be modelled with either a user-defined or Pierson-Moskowitz spectra. Both
unidirectional and multidirectional seas can be modelled.
Ice loads can be applied statically (slowly mooring sheet ice) or dynamically (impacting ice
floes), with the nonlinear constitutive properties of the ice modelled explicitly.
Earthquake loading can be defined using either ground motion records or spectra, for use
in time domain or response spectra analyses, respectively.
Solution methods
Static solution schemes include self-sensing methods. Dynamic analysis methods include
a constant time step procedure using Newmark's integration method and a step by step
integration strategy which automatically adapts the time step to ensure accuracy.
Model generation and and post processing (CAP)
CAP is an environment developed to access information which can help the user understand
a model and its behaviour. This objective is achieved with an interactive,
graphical windowing system through which the user can build or import (from SeaStar,
SACS and STRUDL) structural models, performs queries on those models to highlight key
feature (ie. all flooded members, all pinned joints, etc), run analyses, and post-process
results.
In addition to presenting basic member and structure results (force histories, displaced
shapes, etc), CAP can animate an analysis on screen, while at the same time colour coding
member utilisation ratios and nonlinear events (buckling, hinging, etc). This capability is
extremely useful for helping to visualise and understand the behaviour of a structure
responding to various loading conditions.
The ability to access multiple types of information is demonstrated in Figures 4.4 through
4.6 which show examples of the information provided on-screen for static pushover,
dynamic failure and toppling analyses, respectively. Each of these figures are based on
'snapshots' from animated results of these analyses.
References including the use of include Craig et al
et al et An-Nashif Dolan et Dolan and
Pawsey Nordal (1991) and Bea et al (1988).
Further information can be obtained from: PMB Engineering Inc, 500 Sansome Street, San
Francisco, CA +
11, USA, Fax: 1 415 986 2699.
USFOS
Main fields of application
USFOS is a computer program developed for nonlinear analysis of offshore structures in
steel and aluminium with special reference to progressive collapse. The direct damage
caused by accidental or abnormal environmental loads is assessed in addition to the residual
strength of the damaged structure. The following load conditions can be considered:
Functional loads
Environmental loads (wave, wind, current)
Ship collision
Dropped objects
Fire loads
Explosion loads
USFOS is based on a general, nonlinear continuums formulation for solids. The
formulation is then tailored to 3D frame analysis, based on familiar engineering concepts.
In addition to the nonlinear static analysis, nonlinear time-domain dynamic analysis as well
as eigenvalue analysis may be performed.
Beam element formulation
The beam column is the basic entity of USFOS (Figure 4.7). A coarse element mesh is
used, only one finite element is required per physical member of the structure.
The USFOS beam is valid for large lateral displacementsand moderate strains. An updated
Lagrange formulation is used based on Green strains. The axial force influence on bending
is represented by Livesly's stability functions. Nonlinear material behaviour is modelled
by means of plastic hinges which includes material hardening and the Bauschinger effect.
Spring elements
Linear and nonlinear - single and 2 noded springs are available.
Special features
The basic formulation is extended with special features for load analysis:
Ship collision algorithm including local denting, overall structure deformation and
ship deformation.
Residual strength of tubulars with dents and permanent distortion.
Interaction with local buckling for thin walled tube and rectangular sections.
Tubular joint flexibility and ultimate strength models.
Non structural element option.
Internal, guided pile option.
Fracture criterion for tubulars based on a level 3 CTOD approach.
Degradation of yield stress and elastic modulus at elevated temperatures. Effect
of thermal expansion.
Effect of external hydrostatic pressure on plastic capacity.
Method of solution
An incremental iterative loading procedure employing arc-length iterations to a normal
plane is used. The post collapse behaviour is determined by the current stiffness parameter
and determinant. A bifurcation point analysis is optional. Automatic load step scaling is
offered.
Load avvlication
Proportional and non-proportional loading. Concentrated and linearly distributed loads,
temperature loads, collision loads, acceleration fields. Combination of load cases to load
combinations.
Offshore code checking
Plastic utilisation and von Mises stress check. Member buckling calibrated according to
ECCS code or other. Tubular joint capacity checked according to API and HSE. Ship
collision according to requirements.
Pre-vrocessing and model creation facilities
USFOS may be used as a stand-alone program with ASCII format input files.
The geometry input defined according to the SESAM Interface File Format which may be
generated using the SESAM programs PREFRAME or PREFEM.
Wave and current load data may be generated by the SESAM program WAJAC.
Thermal loads input directly by the user, or generated by the SINTEF program FAHTS.
facilities
Graphic post-processor, XFOS, based on a X user interface. Results
presented as colour fringes (eg. plastic structural utilisation, temperature distribution) on
images of the structure, deformed configurations, XY plots or printed result data tables.
Step by step of visualisation of material yielding, member buckling, force redistribution
etc., up to final collapse of the entire structure.
E
Structure References (Programs) Comparisons
(Note: example publications are cited - where related papers present results of the same or similar analysis, just
one reference is given.)
(a) (b) pseudo structure
axial load
additionalforce in
member caused by
axial deformation
Figure 4.1
Basis of linear superposition method and van de
a) P-M Interaction Surface
for Plastic Hinge Analysis
NODE l
STRESS MONITORING
POSITIONS AT GAUSS
Distributed Plasticity
node
Elastic elements
Figure 4.2
SAFJAC plastic hinge and cubic elements with automatic subdivision of element
LATERAL
DISPLACEMENT
A
LOAD
Upper compression
0
EXPERIMENTAL RESULTS
----- ANALYTICAL RESULTS
LATERAL DISPLACEMENT
Figure 4.5
CAP dynamic failure analysis
Figure 4.6
CAP dynamic toppling analysis
P (Elastic)
Figure 4.7
USFOS basic concepts
ANALYTICAL INVESTIGATIONS
The technical literature now contains a number of papers describing the ultimate response
of frames based on pushover analysis. The basis of the investigations differ, so
representative analyses are described in Section 5.1 before the results are compared in
Section 5.2. Analyses performed to calibrate software against experimental data and
presented in Section 3 are not reproduced, as the intent is to derive new information on the
reserve strength of structures.
2D frames
Idealised 3D jacket
et (1988) interactions
Jacket investigations)
Jacket
Reliability
Eqn 5.1
where p is the rate of unloading following compression brace buckling. The figure
demonstrates the system reserve associated with eccentricity or lack of symmetry which is
inherent in real offshore structures but absent in many simplified models.
and Elnashai
In contrast to the above, the analysis of the frame in this paper involved joint failure prior
to brace buckling. This afforded the structure significant ductility.
and Zettlemoyer
The K-brace subassemblage in Figure 5.6 was analysed using ABAQUS with a shell
element model of the central K joint incorporated explicitly. The same K joint model was
analysed in isolation with typical boundary conditions. Joints with brace angles of 45" and
60" were considered, both with and overlap configurations. It was found that
restraints in the frame enabled the joints to sustain substantially higher loads than
determined from isolated analysis. The results are presented in Table 5.2.
5.2
Comparison of apparent joint capacities within the frame and in
concentric overlap
I 1.11
It was also that the capacity for the joint at that point in the frame was not
strongly influenced by changes to the frame geometry (eg. leg stiffnesses, etc).
The implication from the work is that differences between boundary restraints for joints
tested in isolation and existing in real structures may influence capacity. The work
indicates that 'frame effects' may therefore contribute additional conservatism to
components, which in turn will add to the reserve strength of jacket structures but the
converse could also be the case so further investigation is ongoing.
The experimental programme in Phase of the Frames Project has investigated this aspect
and is being followed by further finite element analyses in Phase IIA (BOMEL, 1992).
Use of the redundancy factor with respect to first yield is demonstrated in this analysis but
the analytical definition of a single plastic hinge is straightforward compared with the
detection of plasticity in a structure in practice.
The USFOS analysis is in close agreement with the SAFJAC results reported by Ward and
Izzuddin under the same loading regime. Ultimate load factors of 1.79 and 1.85 were
achieved respectively, at displacementsof and These values are scaled from
figures in the papers, and therefore agreement may be considered to be very good. The
results of all three analyses are plotted together in Figure 5.9.
As in the first case it was the eventual buckling of the horizontal braces at levels C and D
which brought about unzipping of the bracing system and the major drop in resistance from
steps 9 to l l. It should be noted that these horizontals are typically given nominal
properties within the criteria of design which indicate that the members carry
negligible load under elastic situations.
Design wave loading was increased until the global stiffness decayed. The results for
conventional and grouted pile models indicate the reserve strengths, ie. the ratio of the
environmental load at collapse to the design environmental load. These are presented in
Table 5.5 in which the controlling members in the collapse mechanism are also indicated.
Load deflection curves are not given and loads at which first failure occured are not cited.
Table 5.5
Comparison of collapse and design loads for and jacket
As an illustration, the elastic design utilisation in the struts was around 0.55 for the
South East wave design load. It was clear therefore that significant loading beyond the
design load was required to overcome this implicit reserve at the component level, before
global structural reserves were mobilised. On this basis the levels of reserve capacity are
not unreasonable.
To simulate damage, a series of analyses were performed with members removed. Elastic
analyses revealed the redistribution of loading and in particular higher utilisations for
horizontal bracing which experienced very low loading in the intact state. Furthermore,
nonlinear analyses demonstrated the capacity of the damaged structures which, compared
with the intact capacities, give a measure of residual strength as shown in Table 5.6. The
failure scenarios also reveal the remaining redundancy and the sequence of failures required
to cause collapse.
Table 5.6
Residual strength of damaged jacket structure
The plots of member forces, which are also reproduced in Figure 5.14, indicate that
buckling in the compression brace was preceded by load shedding from another
compression member which was redistributed within the redundant structure. Details of
the plan bracing are not given. From the information presented, comparison between peak
load and load at first failure gives an RF of about 1
Case F was a four leg jacket in water depth for which the plane face frame was
analysed with different bracing configurations, as shown in Figure 5.15. Bracing was sized
from elastic analysis with equivalent interaction ratios for extreme storm conditions.
Design loads and member sizes are not given, nor are failure modes described in detail, so
and RF factors cannot be calculated. However, comparison of the framing reported
in the paper indicates:
The diamond bracing gives the least reserve strength as first member failure limits the
full load path through the structure. (This is not shown in the figure but it is not
certain whether the plateau was plotted or drawn).
The fully X-braced system with horizontals is the most ductile, offering the greatest
reserve capacity as load is redistributed through the bracing.
Substitution of K bracing in place of X bracing causes more sudden failure once the
local capacity is exceeded. This is relevant to the use of K bracing in potential impact
zones or may determine that K bracing should be over-designed to force initial failure
in X panels.
and
The analysis of a platform is presented to illustrate the value of the reserve strength method
in platform optimisations and requalifications. The four leg platform, in water depth,
was analysed for the year design wave. in both the jacket and foundation
were modelled using the element types shown in Table 5.8. The analysis method is not
cited but it is believed to be INTRA.
5.8
Element selection for jacket
Different configurations were analysed as shown in Table 5.9 where the results are also
presented. In all cases failure and the reserve strength were apparently governed by the
first failure cited as loading was shed to the legs which rapidly developed plastic hinges.
Gravity and live loads were applied initially and maintained at a constant level.
Environmental storm loads (wave, wind, current and dynamic loads) were applied
incrementally and increased by a common factor until the global failure mechanism was
developed.
The conclusions from the paper may be presented, with reference to Table 5.9, as follows:
Cases 1 and 2 Clearly the higher the design utilisation of the critical component the lower
is the relative reserve strength ratio.
Cases 3 and 2 Adopting X braces in place of K braces and not affecting structural
weight, a significant increase in reserve strength is achieved and the failure
mode is changed from brittle to ductile.
Cases 4 and 3 Reconfiguration of the X bracing maintains the reserve strength whilst
saving on bracing weight.
Cases 5 and 4 Case 5 similar to Case 4, difference in RSR due to "slightly lower design
unity checks for Case 5".
This is an important demonstration of the role of ultimate strength analysis in optimising
structural configurations for a desirable level of system reserve without compromising
important commercial factors such as steel weight and cost. The point will be discussed
further in Section 6.
Efthymiou and Vugts 988)
Following the verification of INTRA against MARC by the authors reported in Section
5.1.1, the program was used for the nonlinear analysis of a Southern North Sea platform
to demonstrate the use of the limit state assessment criterion with partial factors. The
structure stands in of water with topsides loading of 4700 tonnes and a design base
shear for the North wave of 2150 tonnes. The geometry and elements used in various parts
of the structure are shown in Figure 5.16. Influences of wave loading on member
capacities was accounted for and bending elements were included in parallel with
phenomenological strut elements. A two pass analysis thereby allowed end moment effects
caused by framing action to be accounted for. Three cases were analysed:
Nonlinear structure with linear foundations
Nonlinear structure with nonlinear foundations
Nonlinear damaged structure with nonlinear foundations.
Analyses were performed with 1.15 times the dead loading applied, before environmental
loading was incremented until non convergence occurred at It was then confirmed that
a collapse mechanism had formed. Five wave directions were analysed. The results are
presented in Table 5.10 as in the paper, where the environmental factor at collapse (A,,,,)
is compared with that at which first member failure occurs (X,) to give a measure of the
redundancy factor (RF), where A,,, = X, RF.
Table 5.10
Results from jacket analyses for different structure and foundation conditions
Nordal 991
Three static pushover analyses of the four leg Veslefrikk jacket in 175m waterdepth are
presented using USFOS, FENRIS and (see Section 4). The structure is shown in
Figure 5.17. The ultimate strength analyses were driven by concerns that eliminating the
horizontal X bracing to save weight might reduce the system reserve capacity unacceptably.
Environmental loading was based on 28m wave height, 13.5 second period, year
Stoke's wave with a superimposed 10 year current. Although this specification was the
same in all cases, the structural modelling differed as shown in Table 5.11 below.
Table 5.1 1
Alternative modelling of Veslefrikk jacket
MODELLING COMPARISONS
NW wave design
base shear (MN)
Furthermore, the analyses were performed at different stages of the structure design and
details of the member slendernesses differed. The analyses were not commissioned as a
benchmarking exercise (ie. not with the same brief) and there was no attempt to reconcile
the different results. The analysis was undertaken first and the USFOS and
FENRIS studies were performed later at the detailed design stage (PMB, 1993). The
comparison presented by Nordal is therefore more illustrative of the different approaches
to pushover analysis that can be adopted and of the importance of careful definition of all
parameters, than a fair comparison of the software packages. It is on this basis that the
findings of the paper are presented.
Analyses of the intact and damaged structure were performed. Damage by removal of one
diagonal of the top bay X bracing, simulated ship impact and at simulated incidence
of a dropped object. Collapse was defined by a reduction in global stiffness but the cut off
levels varied in the analyses. In no case was the post failure response traced. Other
differences in the modelling are highlighted in Table 5.11.
The results for the North and North West waves are then compared in Tables 5.12 and 5.13
which follow. The reserve strength ratios relate the environmental load at collapse to the
environmental design load which includes no code load factor. In accordance with
et al a load factor of 1.5 (minimum RSR) is needed to satisfy limit state code
provisions.
Table 5.12
comparisons for under N wave
USFOS FENRIS
Base shear in MN
(load factor)
Damaged:
Base shear in MN
Base shear in MN
Load deflection
Deflected shape
* Damaged at
Figure 5.13
comparisons for under NW wave
USFOS SeaStar
Base shear in MN
(load factor)
at
Base shear in MN
(load factor)
Load deflection
Deflected shape
In addition to the tabulated data, the narrative indicates that the USFOS N wave analysis
shows the peak load is achieved with a factor of = 1.4 on the load at first failure.
After adjustment for base shear, it may be concluded that the USFOS and FENRIS analyses
agree well, whereas the SeaStar analysis for the N wave shows marked differences in terms
of failure mode dominated rather than leg failures) and load factors at collapse (1.8
compared with 3.1). The differences listed in the table are highlighted and the author also
suggests the differences may indicate that the structure is 'well balanced with competing
failure mechanisms'. This conclusion also needs qualificaitonin light of the different stages
at which the analyses were performed and the fact that the basis of foundation modelling
was quite different for the SeaStar analysis. Without resolution of this key factor in
determining the system response it may be concluded that the comparison was invalid.
However, the paper does illustrate the significant value of performing comparative analyses
so that methodologies for structural modelling as well as the performance of different
software can be assessed. The discrepancy between the analyses was large and it is
essential that such factors in relation to jacket performance are verified. For jacket
structures direct calibration is clearly not possible and such comparisons are an important
part of developing a consistency and accuracy across the industry.
First
Yield
Hinge
In addition to the analysis of the jacket structure, a 3 D box frame of similar configuration
to the structure tested by SINTEF (Section 3.1 and Figure 3.35) was analysed using
FENRIS. The member properties and dimensions are detailed in Table 5.16 and differ
from the final test structure.
Table 5.16
3D box frame member properties
Legs S2
S2
Horizontal X brace S2
Vertical X braces S1
S2
Vertical loading was held constant and lateral point loading at nodes (NPL) and in some
cases element loads (EL) were incremented. This differentiated between loading regimes
near the base and top of a jacket structure, respectively. Design loads were calculated to
API and these are compared with the ultimate frame loads for the different brace
slenderness in Table 5.17. Details of the failure modes and redistribution are not given.
Table 5.17
Comparison of reserve strengths for 3D box structure
System Loading Frame Loads RSR
Design Collapse
S1 NPL 44.0 82 1.86
= 97) (no initial imperfection)
NPL 44.0 80.4 1.83
The analyses were repeated with USFOS and good agreement was achieved. In USFOS
one element was adopted per member whereas FENRIS required 108 beam elements and
93 nodal points. Comparative computing times were 1 hour and 2 minutes (see Section 6.3
for further discussion of analysis efficiency). Subsequent USFOS analyses performed by
Soreide et al (1986) for the test structures are shown in Figure 5.23.
Bea, Puskar, Smith and Spencer
The authors outline a general AIM (Assessment, Inspection and Maintenance) approach for
requalifying offshore installations and demonstrate the application to an example Gulf of
Mexico platform. In developing the approach the importance of evaluating platform
capacities and determining the implications of defects was identified. Linear elastic analysis
were shown to be inadequate and potentially misleading giving either conservative or
unconservative predictions. The absence of guidelines or evaluation procedures was noted
but the usefulness of the RSR as a quantified reserve was recognised. Furthermore a
capacity-consequence relation with RSR was introduced as shown in Figure 5.24 and has
been developed further by Bea and Young (1993).
The example platform analysed by the authors shown in Figure 5.25 is a 5 leg (4 corner,
1 centre), fixed drilling platform in water depth which was installed in 1962. The
structure was analysed using with inelastic beam-column elements modelling the
plastic performance of legs, piles and conductors at their load-carrying limit. Braces were
modelled with 'strut' elements to account for buckling in compression or yielding in tension
and the capacities were modified to account for premature punching or tearing at the leg
joint. As the is ungrouted and there are no joint cans this frequently controlled the
capacities. The force deformation of the pile and conductor soil interfaces was modelled
and the authors emphasise the need to account for conductor lateral capacity in order to
model the system capacity realistically. In other words, although such factors are neglected
with a view to conservatism in elastic design, for modelling the nonlinear ultimate response
their influence should be accounted for.
Analysis was performed with environmental loads incremented to failure. Alternative
scenarios were considered, in which:
ALT 1 - platform left as is
ALT 2 - 'repair' missing braces and cracked joints
ALT 3 - repair as 2 and grout legs to piles
ALT 4 - repair and raise deck
Figure 5.25 shows the ultimate responses predicted by the analyses but details of
the failure modes were not given. With respect to the 100-year design load of 1200 kips,
to 1988 standards, the structure in the as is and repaired states, was shown to exhibit
of l and 1.25, respectively. The paper proceeds to discuss the relative
benefit considerations in upgrading the platform.
Shinners, Edwardes, Lloyd and Grill (1988)
In upgrading five Bass Strait platforms, installed in the late 1960s (Figure Esso
Australia recognised that it was not practicable for all components of the platform to satisfy
late 1980s design standards and a target RSR philosophy was therefore adopted. In addition
to more stringent code criteria, the foundation conditions had been found to be worse and
the environmental loads higher than in the original designs.
A study showed that RSRs of recent platforms fell between 1.6 and 2.5, and an optimised
structure could meet current code requirements but have an RSR of only 1 Given more
exact knowledge of actual materials, and fabrication and installation events, a target RSR
of 1.5 was adopted, see also and 1988. Supporting research programmes
were undertaken into soil conditions, K-brace behaviour (Grenda et al, tubular joint
characteristics and material testing.
To determine RSRs, a nonlinear computer model of the jacket was initially loaded with
dead and live loads. Then, increments of the year load were progressively applied as
static steps to collapse. Typically, members with high axial loads and high slenderness
ratios were modelled with nonlinear strut elements; low slenderness ratio members with a
significant component of bending loading were modelled with inelastic beam-column
elements. Strut elements were pin ended and account for post-buckling strength
degradation in compression and strain hardening in tension. The beam-column elements
allow for bending moments at the member ends and account for buckling strength
limitations but do not model post buckling strength degradation. Nonlinear strut and
column elements were also developed to model joint behaviour. Elastic elements were used
for members which were not highly stressed. Special elements were required to model the
sliding of the piles within the legs and preserve lateral compatibility. The deck structures
were replaced by simplified elastic equivalents to reduce computer time. Linearity
assumptions were verified at calculated collapse.
The authors felt that the complexity and multiplicity of failure modes that could occur, as
well as the "judgement required by the analyst", meant that an exact measure of platform
strength could not be achieved although the RSR gave a useful measure.
Analyses for the critical east-northeast direction showed Barracouta and Marlin installations
to be foundation limited, Halibut (right of Figure 5.25) to have equivalent structure and
foundation resistances, and A and B (left of Figure 5.25) to be structure
controlled.
Where structural failure was calculated (and this applied to the strengthened structures as
well) the controlling feature was failure of the east-west K braces and more specifically,
buckling of the compression member or joint within the K brace. The post-buckling, load
shedding behaviour of the K braces leads to an almost "brittle" failure of the jackets. The
shedding of load to adjacent already highly loaded K braces triggered rapid progressive
collapse. The longer diagonally braced north-south column rows were significantly
stronger and did not shed load as rapidly as compression members in K braces.
The need for a increase in capacity was reported for all but Barracouta platforms,
which may be interpreted as RSRs of 0.9-1.0 in the unstrengthened condition.
Alternative upgrading and load reduction options were considered and evaluated using the
RSR approach. Marine growth control and pile struts were recommended although
were shown to be feasible. However, these only became effective when deflections were
large and the dependence on foundation failure to mobilise the action was therefore
not acceptable. The explicit consideration of deflections is unusual however in analyses of
RSR.
Structure
Water depth
Number of legs
Analysis was performed using USFOS by SINTEF and consulting engineers Offshore
Design and Aker Engineering Bergen Modelling was as follows:
Soil was represented by linear springs.
All primary members were given initial imperfections in the dominant
wave direction.
Conductors, conductor framing and topsides were simplified from linear analysis
model.
Tubular joints were not explicitly modelled.
Under cyclic loading the new facility within the USFOS beam element to account for cyclic
plasticity (Eberg et al, 1993) was employed. It is derived from potential energy
considerations and uses a Green strain formulation that allows large local lateral rotations.
The plastification of the cross section is modelled by means of plastic hinges utilising two
yield surfaces in the generalised force space. One surface represents first fibre yield and
the other fully plastic utilisation. Kinematic hardening models are employed for these two
surfaces to account for cyclic material behaviour. The element model was verified by
comparison with a large number of tests on tubular beam-columns and plane frames
subjected to cyclic loading (see Section 3) giving good agreement between measured and
calculated responses provided local buckling or fracture does not occur.
For the pushover analyses, characteristic stillwater loads were applied with the
environmental loads (100 year wind and waves with 10 year current) factored to collapse.
The load factors corresponding to first fibre yield, first member failure (buckling or tensile
yield) and ultimate collapse were recorded enabling redundancy and reserve strength factors
to be derived from the analyses. The loading envelopes for eight directions were reportedly
evaluated but only four or six results are presented in the paper. It is assumed that the least
heavily loaded directions were omitted.
Given the number of load cases in this paper the results are divided into subsections.
Pushover analvsis results
Table 5.19 overleaf, reproduces the results of the pushover analyses. The characteristic
load factors at collapse are presented for comparison with the design value of 1.5 but the
load factors at yield and first member failure are non-dimensionalised with respect to the
collapse load. Within the table the load cases for each structure generating the lowest
redundancy factor (RF) beyond first member failure to collapse are highlighted, as are the
lowest characteristic load factors. It can be seen that these do not necessarily occur for
corresponding loadcases. Nevertheless in all cases the minimum reserve strength (lowest
characteristic load factor 1 is associated with very little margin between first member
failure and collapse. This is demonstrated in Table 5.20 where comparison is also made
with the bracing configuration in the transverse plane. The highest margin between first
member failure and collapse is associated with the only X-braced frame.
Table 5.20
Reserve strength and ultimate response characteristics from pushover analyses by et
Broadside 1 .OO K
Broadside 1.02 K
A2 End-on 1.04 Diagonal
B1 End-on 1.12 X
B2 End-on 1.06 K
B3 Broadside 1.01
strength
The structure exhibited a range of system reserves beyond the design events. This is shown
more clearly in Table 5.21 in which the load factors on the 100 year design load are
divided by 1.5 (product of partial factors 1.3 and 1.15) to give a measure of the capacity
beyond the design event. The basis of the original structure designs is not known so it is
difficult to dra firm conclusions with respect to the absolute RSR values. However, it is
clear that an RSR based on one direction may not be a reliable measure of the structural
system reserve. This can be illustrated by examining the results for structure
in further detail (Table 5.19) and comparing the South (S) and West (W) directions.
Table 5.21
Ranges of reserve strength and redundancy for pushover analyses by et
RF Indicative
The characteristic load factor at collapse for W is 2.81 and the load factor at first member
failure was 2.50 (2.81 0.89). For S at collapse the factor is 2.54 which also corresponds
to first member failure. Thus the load factor at first member failure was slightly greater
than for the W case. In elastic assessment it may therefore by postulated that these critical
members in the W and S load cases would have had similar utilisation. To select a single
load case on the basis of elastic analysis (highest utilisation) for detailed pushover
assessment may not therefore reveal the minimum collapse load factor for the structure as
a whole. Although the W wave would be likely to have generated the higher utilisation,
the ultimate RSR was in fact higher than for the S case. The point is made by way of
illustration and further information regarding the original structure design would be
required for firm conclusions to be drawn in the specific case.
STRUCT LOADING
DIRECTION
Broadside
E
N
0.65
0.79
,,
First
member
0.81
0.96
Table 5.19
Results for structures analysed in pushoverlcyclic investigation
Collapse
FACTOR ON
LOADS AT
COLLAPSE
1.00
1.00
Extreme
9.06
0.98
et
LIMIT LOAD
Amplitude
cyclic
E 0.77 4.17
B2+
I-* Alternating plasticity due to local bending. 'Cyclic capacity' 98% if four members are clamped
+ Results may not be totally representative
112
It should be noted that the account for conservatism in the initial component designs
(eg. member effective lengths) and it is the redundancy factor, RF, the ratio for ultimate
load to load at first member failure, which provides additional information on the system
reserve characteristics.
Although Table 5.21 reveals a range in RSR, the redundancy factors are low and indeed
in only 13 of the 32 cases in Table 5.19 does the margin beyond first failure exceed 10%
of the collapse load, and only seven exceed 20%. The statistic will in part reflect the
configuration of the jackets adopted, nevertheless the structures are reasonably
representative. Focusing only on the lowest ultimate load factor for each platform, the
corresponding margins between first and ultimate failure are in the range 0-12%. However
the load factors at which first member failure occurs are in the range 2.34 to 3.00 which
is significantly higher than the 1.3 to 1.5 range anticipated. The expectation is based on
the limit state factors which are required to produce adequate component designs and the
additional capacity exhibited by the jacket structures is attributed to:
differences in wave theory and current used in design and assessments,
conservatism in design codes and designer's selection of member sizes,
governing design criteria other than environmental loading.
Structure A2 with X and additional cross-bracing in the transverse frames, offers significant
capacity for redistribution (RF, = 1.39). The K-braced jacket B2 also appears to exhibit
good redundancy (RF, = 1.36) for diagonal loading but details of the plan bracing and
the capacity for redistribution cannot therefore be interpreted. However, it is noted in the
paper that first failure occurs in non-critical members which do not play a subsequent role
in the global collapse mechanism. From a system point of view the high RF factors are
therefore misleading although at a local level the component failures may be unacceptable
and require separate assessment.
With regard to determinacy the determinate configurations generally give low RF factors.
Conversely indeterminate systems would be expected to offer capacity for redistribution and
although this is confirmed for end-on loading of structure AO, for broadside of A2 and
partly for end-on loading of A2, insignificant strength reserves are observed for end-on
loading of and B3, and for one end-on loading direction of A2. This is attributed to
the tension members having insufficient capacity to accommodate the load shedding from
the compression braces. Relative member properties as well as bracing configuration are
clearly important.
Although not demonstrated in the paper, it is reported that for non-redundant framing
governed by brace compression failure, there was a dramatic drop in capacity at peak load.
For redundant framing systems, and in particular those having equal number of tension and
compression members, the drop in load was not as pronounced, and the behaviour was
more ductile. In all systems, the energy dissipation in the post collapse range depended on
the yield capacity (or the number) of tensile members, the residual capacity of the buckled
braces and the portal capacity of the legs.
The margin between first yield and first member failure and collapse is also instructive
although it is noted that the USFOS two surface plasticity model underpredicts yield for
components under pure axial loading (Eberg et 1993). In all cases the load factor at
first yield is at or exceeds the target value of around 1.5. Little correlation between first
yield and subsequent member failure can be found, in that early development of a zone of
plasticity does not necessarily lead rapidly to a mechanism of three hinges to precipitate
buckling. Structure B1-East wave gives load factors of and 0.99 respectively, whereas
for structure B2-South wave initial yield at 0.83 is followed by first member failure at 0.94.
No information on the location of initial yield with respect to first member failure is given
to enable further investigation.
Static analysis
et al present details of the pushover structural responses for two of the platform
analysed, and Al. The load deflection curves for three wave attack directions are
shown in Figures 5.28 and 5.29. The load axis represents the characteristic load factor on
the year environmental loads.
Under broadside loading the four compression K-braces in the transverse frames at Level
3 in structure buckle in rapid succession giving a residual capacity plateau associated
with portal action in the legs (Figure For end-on loading (Figure sequential
compressive and tensile failure of the diagonal bracing gives a softening characteristic and
eventual decay in capacity to portal action in the legs. The progressive failures for diagonal
loading (Figure resemble the end-on characteristic.
Structure with K-braced transverse frames exhibit a more brittle response to broadside
loads (Figure It is approximately linear up to a peak load, followed by a sudden
collapse due to extensive brace failure as two compression braces fail over elevation 2 and
the remaining braces at that level fail at the next peak. The platform has heavy legs with
insert piles, so instead of developing a portal frame mechanism between Levels 2 and 3,
further loading leads to failure of the braces above elevation 3. These braces fail at the
next two P-6 'peaks', leading to the failure mechanism shown in the first and second
deflected shapes. From this stage, a portal frame mechanism is developed in the legs over
two elevations (between three horizontal framings) as shown in the third plot.
Longitudinal framing of consists of two diagonal and one X-braced row. The behaviour
under end-on loading is linear up to a peak defined by first member failure. The load then
drops (Figure but to a much higher post-collapse capacity than under transverse
loading. A portal frame mechanism develops in the legs, before further brace failure is
initiated at other levels of the structure.
Under diagonal loading, structure enters into a predominantly transverse failure mode,
and the load-deformation curves (Figure resemble the behaviour under broadside
loading.
Cyclic results
The corresponding results for cyclic loading of the structures were shown in Table 5.19.
Three analyses were performed. In the first the variable amplitude pseudo-storm was
applied. Secondly, constant amplitude cycles were applied to investigate the shakedown
which is in part obscured by the reducing amplitude of the first scenario. Finally the
representative long term cyclic load history incorporating year and extreme storm
loading sequences (Figure 2.6) was adopted. In all but two cases under application of the
cyclic extreme event scenario (Section 2.7) shakedown of the structures occurred, even with
a peak load at 98% of the static collapse value.
Typical results, for structure under broadside loading, are reproduced in Figure 5.30
and are described in the authors' words as follows. indicate the cyclic load
history, with forward and reverse loading on the y-axis and cycle number on the x-axis.
Occurrence of yield hinges are marked on the plot. Some members yield under the first
'100 year' loading (incoming wave), no members yield under reverse year' loading.
Further plastic deformations occur under initial '10, year' extreme storm cycle, but no
member yield under reverse loading. Under the final year' cycles, some yielding
occurs the first time the load is reversed. In the final cycles, the response is purely elastic.
No members experience repeated yielding, and very little cyclic are observed."
The linearity of the global response is evident and the authors report that few cycles (as
shown) are required to verify shakedown or indicate incremental collapse. It is on the basis
of responses such as these that it is concluded that pushover analysis is an acceptable
measure of the ultimate response of an offshore structure.
However, less satisfactory results were obtained for structures A2 under broadside loading
and B3 end-on. There were strong indications of incremental collapse when loads were
factored to the 98% level, with several members experiencing repeated yielding and global
deformations increasing in each cycle. The structures did shakedown for a load factored
to 90% and 93% of the collapse load, respectively.
Taking structure A2 as an example, Figure 5.31 illustrates that significant plastic
deformations (nonlinearity in the response curve) occur within the first cycle of constant
amplitude loading to 98% of collapse. In the static pushover, first member failure was
reported to occur at 72% of collapse. Several members yield under reverse loading, and
in both directions of the remaining cycles. This is indicated in Figure with the
occurrence of yield hinges marked on the plot. Global deformations increase for each
cycle, as shown in Figure 5.3 and Figure 5.3 shows bending moment and axial force
interaction of the critical compression brace as it undergoes severe cyclic yielding.
Shakedown only occurs when the constant amplitude cyclic loading is factored to 84% of
collapse. Under the long-term cyclic load history, shakedown occurred when the cyclic
load was factored to 90% of collapse.
It is emphasised that cyclic degradation is due to the extent of plastic deformations rather
than the absolute value of the external loading above first member buckling. For structure
A2 buckling occurred at 72 % of the collapse load and shakedown was obtained 18 % above
this level. For structure B3 first member failure was at 91 % of collapse but cyclic loads
at only a 2% higher load level could be sustained. The role of alternating plasticity in
exacerbating cyclic degradation is also stressed.
Further discussion of the importance of cyclic loading considerations in relation to static
collapse analysis is presented in Section 6.3.6.
Table 5.23 compares the overturning moments at various stages of the analysis with the
design loads.
Table 5.23
Comparison of jacket capacity with design criteria
Original design
API RP 2A design
API RP 2A design
(8" on diameter for marine growth)
At collapse
In comparison with the original design, the reserve ratio on loads is 3.77 but this is a result
of the low wave height and drag coefficient adopted in the early 1960s.
Adopting the API RP 2A criterion for a wave height and C, of 0.6, reduces the
reserve to and if the presence of marine growth is accounted for this
further reduces to 1.3310.78 = 1.71.However, the allowable design capacity is reduced by
a factor of about 1.48 when code safety factors and differences in material properties
between actual or specified yields are accounted for. Therefore, the system reserve
strength ratio is between 1.39 and 1.16 but it can be seen that first failure occurs below the
current acceptance levels, (ie 1 = 1.38 1.48).
In the adopted by the authors, relating the load at complete collapse to the load at
first failure, the redundancy factor is 1.23 which is quite low but considered to be 'typical'
of such older structures.
J
Environment Most likely Union of any Most likely System
to fail member 1st failure failure path failure
Redundancy = 0.4
I
Table 5.25
Comparison of performance of damaged X- and K-braced structures in Gulf of Maxico environment
Explicit FE modelling of K
Reserve derived from higher
joint capacity in frame than in
5.26 Summary of analytical investigations of reserve strength Simple 2D frames - 2 (Part 2 of 81
STRUCTURE REFERENCE - LOAD-DEFLECTION SYSTEM RESERVE NOTES
(ANALYSIS
METHOD) RSR
Tromans 3.77 (Original With pile failure maximum load factor = 1.0
(USFOS) design load)
2.22 (API)
RELIABILITY
et DemonstratesMonte Carlo analysis alone may
not reveal correct probabilities for different
failure modes. Combinationof Level
reliability analysis proposed.
5.2.4 2D versus 3D
A number of pushover analyses have been undertaken in which plane frames have been
considered to investigate aspects of jacket performance (eg. Piermattei et al, Jacobs and
Fyfe). In the first example, 3D analyses were undertaken subsequently, but in the work
presented elements had been between the two analyses such that direct comparison
cannot be made. The best example is therefore given by the simple frame analyses by Pike
and Grenda where eccentric loading was redistributed into the alternative bracing plane as
failure in the first was precipitated. Whereas in 2D analysis first failure dictated ultimate
load, 3D redistribution enabled frame reserves of more than 30% to be exploited. This
illustrates the principal of 3D frame action within a jacket and the dependence on adequate
plan bracing to transmit the loads. This was also demonstrated by Lloyd. If sufficient
bracing is not provided, the ultimate response may still be governed by first member
failure.
3D action depends also on the direction of incident loads. In practice, the direction with
the lowest RSR generally forms the focus for descriptions of failure and investigations of
redundancy (Piermattei et al). A systematic investigation for a loading rosette with base
loads related to directional sea states may be more meaningful, together with correlation
to elastic component (eg. Tromans and van de Graaf, et al, etc).
IDEALIZED 3D JACKET
Not included Members removed to evaluate redundancy influence
Lloyd Influence of grouted leg piles on jacket response included Members removed - asymmetry
Lloyd Not included Members removed and weight to achieve
JACKET STRUCTURES
Jacobs Fyfe Not included. Members removed.
Nonlinear springs between piles and soil. Instances of X
failure where compressive skin friction exceeded,
et al X X
Holm et al X X
Nordal et X Reliability for damaged state
means nor included in
The role of comparative analysis
The analyses presented in Section 5 have revealed important aspects of jacket reserve
capacity. However discrepancies between base design criteria the availability of key
details have limited the comparisons that can be made. In contrast the few comparative
analyses that have been presented (see Table 4.5) have been most instructive.
Such comparisons enable not only the performance of different software packages to be
assessed but also provide unique opportunities to validate different approaches to modelling
and analysis. This is particularly important given the current absence of physical data for
complex 3D structures against which software can be benchmarked, and the increasing
reliance that is being placed on the calculation of by the offshore industry. A
principal recommendations from this review (see Section 7.2) is that further comparative
analyses should be undertaken and published as a means to develop the education process
in this important state-of-the-art technology area.
ASSUMPTIONS
E 29.000
36
C
0.8
P., of
P, F, Area
-014
DESIGN TRADE-OFF
OPTION lI HORIZONTAL BRACE
21 HORIZONTALBRACE
Figure 5.1
Two bay X-braced frame analysed by Pike and Grenda
NOTE:
FRAME l
0 -2 -4 -6 -8 -
BRACE UNLOADING RATIO
Figure 5.2
Parallel K-braced frames showing influence of eccentric loading and rate of brace unloading p on
reserve strength (Pike and Grenda)
JOINT 1
SECTIONS
l
JOINT 2
Figure 5.3
jacket frame analysed by et to illustrate joint flexibility effects
SHEAR FAILURE (BRACE)
FAILURE (BRACE)
RIGID
FLEXIBLE JOINTS
Figure 5.4
of joint capacity on response identified by Ueda et
: hinge
Flexible Frame
(m)
Figure 5.5
Analysis of a jacket frame with flexible and rigid joints by Elnashai and Gho
Figure 5.6
Frame models analysed by and showing differences in joint capacity and
post-ultimate axial-moment interaction both within the frame and in isolation
t U
REFERENCE
IS
6 E LE M ENTS
- ELEMENT
0 20 0 3% 0 0 4% 0 0
DISPLACEMENT (m)
Figure 5.7
Jacket frame analysed by et using MARC and
IS LATERAL LOAO
TOTAL X LOAO
TOTAL Y TONNES
a . INTACT
.
0.0, . , . , . , . , .
Figure 5.8
Analyses of jacket frame by Ward and lzzuddin and et
- MARC
INTRA
SAFJAC
SAFJAC
0 0.5 l
LATERAL LOAD
Figure 5.9
Comparison of jacket frame analyses by et IMARC, INTRA),
Ward and lzzuddin (SAFJAC) and et
MEMBER INELASTICITY
STRUT
@m-
@
Figure 5.10
Comparison of energy absorption ductility for 2D jacket frame analysed by Gates et
BROADSIDE LOADS
1.2 elastic
TRUSS 3.8 .79
Figure 5.11
Redundancy calculations for Gulf of Mexico structure due to
ROW l A
Figure 5.12
and X-braced jacket
CASE A
---- CASE B
--------- CASE C
I
RESIDUAL STRENGTH
Figure 5.1 3
Four-legged jacket used in redundancy study by Lloyd
V EL. -
V EL. -
;-- EL. - 41
,
30
o , I I
o 40 80 o
Figure 5.14
Intact jackets pushover analysis by Jacobs and Fyfe
Figure 5.15
Brace configuration study covering X, K and diagonal bracing (Jacobs and Fyfe)
l
Figure 5.16
Jacket analysed by et
Figure 5.17
Analyses of Veslefrikk jacket report by Nordal
FRAME 1 FRAME G
01 02 0.3 07 08 09
Lateral Displocernent (meters)
Figure 5.18
Nonlinear analysis of four-legged Goodwyn concept et
Lateral Displacement (meters)
Figure 5.19
Nonlinear analysis of eight legged Goodwyn concept et
Figure 5.20
Failure sequence for the A platform et
Figure 5.21
Analysis of 3D jacket structure et
Figure 5.22
Eight leg jacket analysis by Moan et
Figure 5.23
3D box frame analysed by Sordde et
Figure 5.24
Capacity consequence evaluation due to Bea et
Figure 5.25
Load displacement curves for example jacket for AIM alternatives et
AUSTRALIA
LOCALITY
- OTHER PLATFORMS
PIPELINES
Figure 5.26
Esso Australia platforms for upgrading using RSR approach et all
Structure
Figure 5.27
Structures analysed in investigation, to scale et
Lading
Figure 5.28
Pushover responses for structure analysed by et
Figure 5.29
Pushover responses for structure analysed by et showing deflected shape
""m*!
Load
Figure 5.30
Cyclic analysis of structure under broadside loading et
Figure 5.31
Cyclic analysis of structure A 2 under broadside loading et
A m
t
Structure failure surface
Figure 5.32
South Pass SP62-B jacket and van de Graaf)
Environmentalbad
Row A 1 h3
Elevation views of
Load lateral
Load deck
Figure 5.33
Gulf of Mexico platform used for hindcarting analysis (van de and
(leg)
displ.
Figure 5.34
Failure modes in reliability analyses by Edwards et
50
3.0
17, WAVE :
stress
bar 2 2 3
to
Compression yield
of bar 244
\ X
240 120 60
Figure 5.35
Influence of yield on system behaviour demonstrated by Holm et
FORCE
SEMI - BRITTLE
DEFORMATION
Figure 5.36
Ductilelbrittlelsemi-brittle response of member considered by Nordal et
MEMBER
GROUP DIA. THICK.
KB2
LG2
LG4
Figure 5.37
K-braced variant of Figure 5.12 adopted by Nordal et
6 . DISCUSSION
Influences on RSR
For jacket structures, measures of RSR are often presented in terms of base shear or
overturning moments. This needs qualification as the same storm loading from different
directions will generate different base shears and overturning moments. Again, as a
relative measure, the approach may be satisfactory but in absolute terms results should be
interpreted with caution.
The importance of loading direction and correlation between elastic utilisation and ultimate
RSR is not explored in the literature. However, it is particularly important in the
assessment of jacket structures. Based on elastic analyses for a rosette of wave loadings,
the critical direction and corresponding wave height giving the greatest utilisation in any
component may be found. In incrementing the loading to collapse, the structure may have
significant capacity for redistribution beyond first failure giving a high RSR. Subject to
loading from a different direction, first failure may be at a higher load factor (ie. the
critical component has a lower elastic utilisation), but this may trigger a rapid sequence of
failures such that only a low RSR exists. The peak capacity may also be less. Clearly the
critical load case under elastic design criteria, does not necessarily correspond to the worst
case in terms of reserve strength. Demonstration of this fact is required in the open
literature to ensure that adequate analyses are performed for jacket assessments.
Analysis for a rosette of environmental loads (eg. van de and Tromans, 1991) is to
be recommended but may be considered too costly or time consuming. However, if a
reduced analysis approach is to be followed, it is clear that it needs to be established on a
firm and rational basis.
6.3.4 Role of tubular joint failures
Most test data are for specimens where members have been the critical components.
Almost all the analytical predictions of reserve strength for jacket structures have
considered critical members and rigid joints have been assumed. In fact, many older
structures were constructed without joint cans so that the tubular joints are highly utilised
and in some cases are 'critical' components. Even where cans are present, once storm
loads are factored in the determination of RSR, the capacity of these joints may be
approached and significant nonlinearity in the local response is to be expected. Joint
checking criteria and modelling capabilities within pushover analysis programs are therefore
required. It should be emphasised that tubular joint criteria are based on the peak loads
achieved in isolated tests and are not related to the limit of linear behaviour or restraining
effects of chord fixity or continuity.
The platforms analysed by Bea et and Shinners et (1988) were older structures
without joint cans and due to limitations of the available software these authors simulated
tubular joint failure by modifying brace capacities in the first case, and by developing
specific nonlinear strut and beam-column elements in the second. Since 1988 however, the
published results have generally focused on limiting member behaviour and advances in
modelling tubular joint failures are not generally reported.
Depending on the joint configuration, nonlinear behaviour may be exhibited at 40% of the
peak load. It is therefore certain that at the ultimate load levels being predicted for jacket
structures, nonlinear joint behaviour will have commenced. The implications from ignoring
this may be significant. As demonstrated in the SCI and BOMEL tests, softening in the
response of a joint would limit incoming brace loads and so protect the members
themselves from failure. Loads shed to alternative loads paths may generate failures
elsewhere. It is clear therefore, that nonlinear joint behaviour may not only alter the
capacity of a structure but may also affect the sequence and mode of failure predicted.
Indeed, an ultimate strength analysis needs to incorporate all nonlinear aspects of the
behaviour if reliable predictions are to be made.
A problem with joint behaviour is the lack of information on ultimate response
characteristics, in addition to capacity predictions, from the available database. Not all
analysis programs have the explicit facility to model joint behaviour althcugh modifications
to phenomenological strut models may be made to give an effective response characteristic.
Other programs (eg. SAFJAC) enable force-deflectionand moment-rotation characteristics
to be defined, but these need to be specified for each configuration.
A detailed evaluation of the potential influence of joint behaviour on the reserve strength
of jackets needs to be made, both to ensure that confidence can be placed in predictions,
and to determine whether additional information beyond that currently available is required,
eg. for multiplanar joints. Consideration of the influences of thickened joint cans on the
effective buckling lengths of incoming members may be required in some instances.
Consider two structures designed to the same code with the critical member in each
having the same utilisation. Under extreme loading it may be anticipated (based on
idealised assumptions) that these members would fail at the same global load factor.
If the first structure were non-redundant the system capacity would fall. In a redundant
system plastic deformation of the critical member might take place as loads were
redistributed until the ultimate strength was reached at a higher load factor. Figure
6. l a illustrates the cases.
Both structures may be expected to achieve shakedown around the first member failure
load, and therefore perform satisfactorily at the same absolute load level. Although the
subsequent plastic deformation in the second structure may cast doubt on the validity
of the ultimate capacity because of the potential inability to achieve shakedown, this
may be of less concern. Any capacity beyond the initial failure load level is a bonus
with respect to the first design which would undergo a static collapse if the load were
exceeded.
2. Reassessment of older structures
A potentially greater concern may be in reassessment of structuresdesigned to outdated
codes with varying levels of safety factor and different relative component capacities.
Figure 6. l b gives two examples where the structures exhibit the same ultimate static
capacity but the first behaves in a brittle manner without the redundancy and
redistribution of the second. It is supposed that both structures are now subject to the
same design storm and, as shown, both can achieve the same global load factor.
In the second case the level of first member failure suggests the structure would fail
to meet current component based criteria and this is a practical case in which pushover
analysis may be used to demonstrate system adequacy. However, it is here that cyclic
considerations may be important. If the second structure relies on large plastic
deformations to achieve the peak load, it may be that shakedown could only be
achieved at a lesser load. In this case the pushover load alone could be misleading and
give an unconservative measure of system performance. The more brittle response
could perhaps be accepted more readily without further cyclic investigation.
3. Design based on global resistance
The responses in Figure 6. can also be used to illustrate the need for both component
based and system based considerations in design. Were two new structures to be
designed on the basis of a system pushover load factor alone, the two responses shown
would be equally satisfactory, despite the concerns from a cyclic loading viewpoint
expressed for the second case under item 2 above. By having a corresponding
component based criterion the second structure would only be acceptable if first failure
were achieved at a higher load level than that shown, thereby potentially increasing the
system load at which shakedown could occur.
The above examples are simplistic, nevertheless they serve to illustrate the interaction
between cyclic and static pushover behaviours which need to be considered in the
application of these techniques and the development of a philosophy for structural
evaluation.
Beyond the high stress low cycle plasticity effects which are being modelled, future
consideration also needs to be given to low stress high cycle fatigue which may result at
points of stress concentration associated with component failures, fracture potential and
tubular joint performance under cyclic loads.
The need for more data extends also to the assessment of inertia and near failure dynamics
and strain rate effects which may modify the calculated. This is an aspect
requiring further investigation if the realistic performance can be accurately modelled.
6.4.2 Validation
Validation is a prime concern in the application of nonlinear software to jacket structures.
Closed form solutions are available for calibration on a component basis. Two 2D test
results are available for benchmarking purposes (HSE, 1993). It is only
comparison between analyses using different programs and by different users of the same
program which offers any opportunity for validating the complex nonlinear redistribution,
failure sequences and modelling approaches for 3D jacket structures. However in many
cases the analyses are sensitive to input assumptions and there is a clear need for data on
component characteristics to be available as well as controlled tests for
calibration. A further phase of Frames Project is planned (Bolt et al, 1994) to
encompass representative collapse tests of a 3D structure to provide a basis for more
rigorous benchmarking.
Whilst this uncertainty remains, it must be necessary to impose a partial factor on the
predictions. This will result in a higher requirement for the target RSR than will be
required as development in modelling ability and experience improve 'confidence'.
Load Load
Factor Factor
First Member Failure
Load Load
Factor Factor
Figure 6.1
Comparisonof and ductile responses for consideration of cyclic effects on achievablecapacity
7. CONCLUSIONS AND RECOMMENDATIONS
The ability to predict the reserve strength of jacket structures is now of considerable
importance to the offshore industry. There is a requirement to extend platform operating
life despite more onerous loadings and more stringent code requirements than at the original
design stage. Furthermore, risks of extreme events which cannot reasonably be resisted
elastically, have been identified and adequate system reserve is therefore a necessary
consideration in configuring new jackets. Redundant structures have an inherent ability to
redistribute loads as plasticity occurs such that first component failure need not be
synonymous with structural collapse. It is this reserve strength which must be utilised to
demonstrate that loads beyond the original design scenarios can be sustained safely.
This recognition of the importance of reserve strength technology has been met by the
development or adaptation of a range of nonlinear software to perform the collapse analysis
of jacket structures. These embody different approximations and numerical devices with
a view to ensuring that the complex nonlinear problems can be analysed efficiently and to
sufficient accuracy. Despite calibration at the component level, benchmarking against the
available test data identified in this review is recommended. Furthermore, without 3D test
data for direct comparison, recourse should be made to comparative analyses so that by
resolving discrepancies, more may be understood about the complexities of nonlinear
responses.
Beyond the variability in software capabilities, is the methodology for performing analysis
and evaluating reserve strength (RSR). The review has indicated that a simple approach
is generally adopted, whereby stillwater loads are maintained constant and environmental
loads are factored - the factor on the design load at the peak determines the RSR. In
discussion, this review has suggested that although providing a relative measure of system
reserve, the RSR has little physical relevance and more meaningful assessment strategies
are proposed, where appropriate. Furthermore, the ability of a structure to withstand
extreme loads needs to be assessed not only in relation to capacity but in terms of deflection
criteria, the condition of the structure once the loads have abated and the subsequent
structural performance.
This discussion is relatively new and many of the issues raised remain to be resolved. The
principal recommendation from this report is therefore that a procedure for modelling and
evaluating structural reserve should be devised and verified against a series of comparative
cases so that it may form the basis of agreement with industry.
This review has focused on reserve strength and the lessons from static pushover analysis
that can be applied to the assessment of the performance of jacket structures. Specific
conclusions from the present review are detailed in the following subsections and are
followed by recommendations in Section 7.2.
7.1 CONCLUSIONS
Test data
0 Test data are available demonstrating various combinations of member and tubular joint
failures for 2D and 3D structures. Four key programmes relate to offshore structures,
namely Popov, SCI, Grenda et and BOMEL.
As the data pass into the public domain, they provide a valuable and essential basis for
benchmarking software.
The reserve strength from the alternative load paths through X-braced panelling is
demonstrated in contrast with the lack of redundancy in K-bracing (or single diagonal)
bracing.
Member failures have correlated well with predictions and have provided valuable
evidence for effective lengths.
Ductile tubular joint failures have protected load paths affording considerable ductility
to the global response and enabling significant system reserves to be developed.
Differences from tubular joint responses predicted from isolated tests have been
observed in the frames due to boundary conditions and the combinations of brace and
chord loading which occur associated with frame action.
3D test data are inadequate as the available data relate to non-offshore configurations
and structures with initial
Numerical data
Simple models have been shown to be valuable in elucidating the mechanisms
underlying nonlinear system responses.
Comparative analyses for different bracing configurations can readily identify efficient
material distribution to improve structural system reserve and illustrate the significance
of relative member properties and redundant members in the process of redistribution.
A wide variety of software programs are adopted by industry (eg. ABAQUS, EDP,
FACTS, INTRA, KARMA, RASOS, SAFJAC, USFOS) based on variations
of four methods
- the finite element method
- phenomenological models
- polynomial beam column modelling
- structural unit method.
Descriptions on a common basis are not readily available and the review has
reproduced descriptions provided by the software developers.
Comparative analyses have been presented revealing different failure modes and loads
for the same problem. Initial concern is allayed by differences in foundation
modelling, but this reinforces the need for comparative analyses where benchmarking
for jacket analyses is not possible.
In specific cases, jacket analyses identify the importance of foundation characteristics
and the application of realistic load redistributions. However, no systematic assessment
or sensitivity studies are presented.
Few analysis cases are presented where tubular joint failures play a part. The
conclusion is that this reflects the modelling complexity and paucity of data (as for
foundations) and not the known situation amongst older offshore structures.
The pushover analyses presented represent a range of purposes such as design
optimisation, achieving target RSR at the design stage, RSR evaluation as part of
earthquake requalification, hindcasting etc.
In the absence of data and for simplicity, damage is generally modelled by removing
members which may or may not be conservative for the global system response.
Collapse analysis is generally performed by holding stillwater loads constant and
incrementing environmental loads.
Magnitude of RSRs
Comparison of reported RSRs is difficult because the basis of design loads and the use
of WSD or LRFD approaches are a source of confusion.
Target RSRs of 2.0 are reported for intact structures compared with an implicit reserve
for the design process of around 1.5. The above reservations apply.
In many instancesa significant reserve exists beyond first member failure but in several
of the platform analyses presented first member failures triggered collapse.
7.2 RECOMMENDATIONS
On the basis of the above conclusions, the following key recommendations can be made:
Data. Additional data on key factors contributing to the nonlinear response of jacket
structures are required in the areas listed above. Evidence from the tests and analyses
reported is that foundations and tubular joints are particularly significantand urgently
require consideration given the configuration of older and foundation
conditions.
Analysis. Information and criteria on which to select appropriate analysis tools for
specific situations are required. Benchmarking against test data will be a valuable first
step. Comparative analyses of the same structure using different programmes and the
approach of different analysts will follow. In test data for jacket type
frame structures and joints should be generated to provide a rigorous test for
Greater confidence in the analytical capability and accuracy will lead to
reduced RSR requirements and greater economy.
methods. Current approaches to load application are simplified giving an
indicative measure of reserve strength but without modelling all aspects of loading
regime. A more rational methodology for incrementing loads to collapse should be
devised and validated, including for example guidance on the treatment of lateral
member loads, the use of mean (or characteristic) component capacities, selection of
orientation, the need for sensitivity studies, etc.
Acceptance criteria. Linked to analysis methods is the need to define target in
relation to both WSD and designs. criteria should reflect the purpose of the
analysis and should extend beyond the RSR to
the consequences of overload, deflection criteria,
vulnerability of key components to other risks etc, so that the continued safe operation
of the beyond the extreme loading event may be considered. Acceptance
criteria need to be laid down and in conjunction with the formalisation of
analysis methods noted above.
This report has reviewed the current state of practice embodied within the literature
regarding reserve strength technology. Tools are now readily available to perform collapse
analyses but the complexities of nonlinear analysis demand that the results be investigated
and explored to ensure that meaningful and accurate predictions of the ultimate response
of structures underlie the numbers generated. Furthermore for many packages,
development is ongoing reflecting the complexity of the nonlinear situation that is being
modelled.
REFERENCES
H
Assessing fitness for purpose of offshore platforms - Risk management, maintenanceand
repair
ASCE Journal of Structural Engineering, Vol 120, No 12, December 1994
BEA, R G
Structural reliability: design and requalification of offshore platforms
Reliabilityof Offshore Operations: Proceedings of an International Workshop. E
NIST Publications 833, 1992
BEA, R G
Development and verification of a simplified method to evaluate the on and
capabilities of steel, template-type platfonns
Energy and Environmental Expo, Houston, 1995
AL-BERMANI, F G A KITIPORNCHAI, S
Nonlinear analysis of transmission towers
Eng. Struct., Vol 14, No 3, 1992.
AN-NASHIF, H
modelling for offshore structures
International Offshore and Polar Engineering Conference, San Francisco, 1992
BLUME, J A
A reserve energy technique for the earthquake design and rating of structures in the
inelastic range
Proceedings 2nd World Conference on Earthquake Engineering, Tokyo, 1960
BOLT, H M
Results from large scale ultimate strength tests of K braced jacket frame structures
Offshore Technology Conference, Paper No. OTC 7783, Houston 1995
CA
Structural reliability - some contributionsto offshore technology
Offshore Technology Conference, Paper No. OTC 7753, Houston 1995
CRAIG, M J K and KA
Assessments of high consequence platforms - issues and applications
Offshore Technology Conference, Paper No. OTC Houston 1994
DE, R
Risk analysis methodology for developing design and assessment criteria offshore
structures
Offshore Technology Conference, Paper No. OTC 7755, Houston 1995
W A and IBBS, C W
Assessment and of offshore production structures
Workshop Proceedings, New Orleans, 1993
DYHRKOPP, F G
Minerals Management Perspective of platform assessments
Energy and Environmental Expo, Houston 1995
EBERG, E, AMDAHL, J, T and HEKKELSTRAND, B
Integrated analysis of offshorestructures subjected to
Conference on Design against Accidental Loads as part of the Offshore Safety Case,
London, 1992
ECCS
European recommendations for steel construction
FISHER, P J
Some practical considerations for reassessing offshore
Requalification and Reassesment of Offshore Installations Workshop, Institution of
Mechanical Engineers, 1994
HOLNICKI-SZULC, J GIERLINSKI, J T
Structural modifications simulated by virtual distortions
Int. J . for Num. in Eng., Vol 28, 1989
KALLABY, J and DN
Inelastic analysis of offshore platforms for earthquake loading
Offshore Technology Conference, Paper No OTC 2357, Houston, 1975
KARAMCHANDANI, A and CA
An event-to-event strategy for nonlinear analysis of truss structures I
ASCE Journal of Structural Engineering, Vol 118, No 4, 1992
KARAMCHANDANI, A and CA
Reliability analysis of truss structures with multistate elements
ASCE Journal of Structural Engineering, Vol 118, No 4, 1992
RW
Ship collision analyses for the Island steel platforms
Offshore Technology Conference, Paper No OTC 7 1 , Houston, 1993
LLOYD, J R
Design strategy for redundant space frames
1982 Structural Stability Research Council Annual Technical Session, New Orleans, 1982
LLOYD, J R and WC
Reserve and residual strength of pile founded, offshore platforms
In The role of design, inspection and redundancy in marine structural reliability, National
Research Council, National Academy Press, Washington, 1984
MARSHALL, P W
Strategy for monitoring, inspection and repair for offshore structures
Behaviour of Offshore Structures Conference, London, 1979
MARSHALL, P W
Failure modes for offshore platforms -fatigue
Behaviour of Offshore Structures Conference, Trondheim, 1976
MARSHALL, P W
Screening offshore platforms: previous approaches and further thoughts
Proceedings of Civil Engineering in the Oceans V, ASCE, Texas, 1992
MOSES, F
System reliability developments in structural engineering
Structural Safety, Volume 1, 1982
NORDAL, H
Application of ultimate strength analysis in design of offshore structural systems
In Integrity of Offshore Structures - 4, ed. Faulkner, D, Glasgow, 1991
G
Non-linear structural dynamics by the influence method. Part I: Theoretical
considerations
International Offshore and Polar Engineering Conference, San Francisco,
G
Non-linear structural dynamics by the influence method. Pan 11: Application
to offshore collapse
International Offshore and Polar Engineering Conference, San Francisco,
G, and TROMANS, P S
Nonlinear reassessment of jacket structures under storm cyclic loading. Pan -
Representative environmental load histories
Offshore Mechanics and Arctic Engineering Conference, Glasgow, 1993
P G and H
Reserve strength analyses of offshore platforms
Offshore Southeast Asia Conference, Paper 88179, February 1988
ZHANG, S K
The failure load calculation of offshore jacket by incremental limit analysis
Offshore Mechanics and Arctic Engineering Conference, Japan, 1986
APPENDIX A
REVIEW UPDATE
AUGUST 1993 - MAY 1995
INTRODUCTION
The main text of this Review was completed by February 1993. The document was
expanded in August 1993 specifically to include a series of four papers et al;
and Tromans; et al; Eberg et 1993) examining the validity of static
pushover analyses to evaluate ultimate structural response characteristics in a cyclic storm
loading environment. In reformatting the review in anticipation of publication in January
1994, reference was included to the first draft of a Section 17.0 to API RP 2A-WSD for
the assessment of existing platforms which was first published in December 1993. This
appendix has been added in May 1995 to describe the principal developments in relation
to industry's understanding of ultimate system strength and its application of the technology
in the intervening period.
Hurricane passed through some 3,000 offshore structures in the Gulf of Mexico
in August 1992 and provided the opportunity for analyses to be performed to
evaluate loading and resistance models against the failurelsurvival experiences. It was not
until 1994 that the findings could be consolidated and these recent results are discussed both
collectively and in relation to individual structural performances in Section A2.
This body of Hurricane information also enabled the API Task Group, TG
drafting Section 17.0, to develop experience based criteria for the future assessment
of existing platforms in the region. Recognition is given in Section 17.0 to the value of
ultimate strength analysis and a reserve strength ratio (RSR). However it is important that
the basis of the US provisions are understood so that appropriate criteria can be derived for
the acceptance of RSR elsewhere. Section A3 of this Appendix sets down the background
to the definition and use of RSR in Section 17.0, and details the anticipated timescale for
adoption in API RP 2A and the standard.
To validate the approach to assessment in Section 17.0, MMS initiated a JIP encompassing
trial applications of the procedure and a benchmark ultimate strength analysis of a specific
platform. The findings give some insight to the confidence that can be placed in the
application of the reserve strength technology. These and the results from the HSE
benchmark against the Frames Project test data are reviewed in Section A4.
Finally Section A5 revisits the discussion, conclusions and recommendations presented in
Sections 6 and 7 of the main report. Several of the key areas identified have been
addressed but others remain unresolved and the conclusions in this Appendix underlie these
considerations for the future safe application of the reserve strength technology.
The discussion in Section 6 of this Reserve Strength Review identified key issues related
to the evaluation of system capacity and the appropriate exploitation of the reserve strength
technology. Between 1993 and May 1995 a number of papers have provided new
information either illustrating or helping to resolve some of the problems faced. For this
report to reflect the state-of-the-art at the time of publication, a selection of these papers
is reviewed.
Most of the work has been based on the use of numerical analysis techniques (discussed in
Section 4) for individual platform investigations, in the manner of results already presented
in Section 5. However, in some cases important links have been made to the experimental
database, presented in Section 3. Furthermore the accumulating industry experience has
been consolidated in a number of important references where direct comparison of reserve
strength is made on a common basis across a range of structures. All these areas are
covered in the reviews which follow. Table A2.1 cites the references in order of the
reviews and identifies the key contributions in relation to this review of reserve strength
technology.
A2.1
Summary of key references 1993 - May 1995
Reference Key issues
van de Graaf (1993.4) Dependenceof RSR on environment and design code basis.
Plane frame
2D jacket
North Sea Jacket
broadside 2.90 3.28
end-on 2.18 2.35
The conservatism attributable to the K factor can be seen clearly by comparing the first and
second results columns. By comparing the second and third columns it can be seen that
appropriate selection of effective length factors can give a more meaningful representation
of member failure.
The effects are studied more extensively in the paper and two further conclusions may be
cited. Firstly, the load level at first member failure in these structures is little influenced
by the potential for plasticity to develop elsewhere in the structures. Secondly, the
representation of member collapse by a refined K factor is satisfactory when the maximum
moment occurs near as this coincides with the simplifying assumption of hinges
forming only at member ends in USFOS (see USFOS description in Section
4). For conditions of double curvature the largest discrepancy (21 %) was found.
Completing the investigation, the effects of initial imperfection on first member failure and
system collapse loads are reported. Table A2.3 shows the influence of a 0.5%
imperfection.
Table A2.3
Percentage reduction in collapse loads for a 0.5% imperfection
It can be seen that for these X-braced structures, where system failure is governed by more
than one member, that there is relatively less influence of imperfections on collapse strength
than individual member failure. For end-on loading of the jacket structure two compression
braces shed load onto one tension brace and the influence on system collapse load is
therefore more significant.
This paper demonstrates the value in developing a wider appreciation across the divide
between linear elastic design analysis and nonlinear pushover analyses. Furthermore it
highlights the potential significance of imperfections on both component and system strength
depending on the structural configurations and redundancy.
Botelho, Petrauskas, and Kan
This paper is the first of several in this Appendix dealing with analyses undertaken
following the passage of Hurricane through some 3000 offshore structures in the
Gulf of Mexico in August 1992. Ten major platforms were toppled directly by the
hurricane which involved sustained winds of 140 miles per hour. Twenty-five satellite
installations (including caissons) were also toppled with significant damage incurred by a
further 26 major and 140 satellite platforms.
This event provided an important opportunity to validate the use of the ultimate strength
analysis techniques both at the individual platform level, hindcasting collapse, damage or
survival, and collectively to provide a basis for future risk management of the many
structures in this hurricane prone region (see et al, 1994 reviewed below).
This specific structure, ST was an 8 pile platform installed in a water depth
in 1958 having been designed to the then standard 25 year criteria. However, the deck had
been raised in 1991 (Figure A2.1) in anticipationof significant loads should waves encroach
the deck in extreme weather. Nevertheless the platform was completely toppled by
Hurricane .
Pushover analyses were performed by the authors using CAP (Section 4) and some details
of the modelling approach are presented. Effective length factors (k) were set at 0.65,
except for X braces which were modelled as two diagonal struts, eliminating the X joint,
with k equal to 0.325. For face frame K nodes the panel capacity was deemed to be
governed by joint failure and the member characteristics were chosen to simulate the initial
failure characteristics. The authors point out the shortcomings of this approach in that
redistribution of member loads cannot take place in the event of joint failure which may
sever the brace-chord intersections. Furthermore the analyst must determine whether joint
or member behaviour will dominate for a particular wave direction analysis.
The importance of this can be seen in the results in Table A2.4 which identifies the
significance of joint failure for all three wave attack directions.
Table A2.4
Pushover analysis results for 130 "A"
End-on 1,550
Pile Hingesl
Braces
Buckledl
The load deflection response for broadside loading in Figure A2.1 also reveals a ductile
softening characteristic, whereas evidence from the BOMEL frame tests (Section 3 and
Bolt, 1995) indicates that gap K joint failure may precipitate rapid load shedding requiring
redistribution of forces throughout the structure. In presentation at the Offshore
Technology Conference (Botelho et al, photographs were shown of platforms
recovered after Hurricane with complete severance across the gap region of K
joints. Figure A2.2 shows the line of frame action driving shear failure in the gap K joints
in frame tests (Bolt, 1995). This mode of failure was somewhat unexpected in
the frame tests but the complete severance across joints in Hurricane confirmed
the value of the tests in highlighting frame action and potential failure modes.
The table indicates redundancy factors (failure nonlinear event) of 1.7, 2.3 and
1.6 for ST in the three directions. These are substantial values which may not be
achieved were it possible for post-ultimate response of the K joints to be modelled.
Comparison of the base shears from the Hurricane in the final two columns of
Table A2.4 shows the need for the uncertainty in various parameters to be accounted for.
With a COV of 0.25 the probability of failure for ST under the diagonal wave was
some 54%. However, whilst this tied in with the load at failure of the platform,
it should be noted that ST located in the same area survived without damage
despite a very high probability of failure calculated on a similar basis (Petrauskas et al,
1994).
The paper highlights the considerable importance of a rigorous joint modelling capability
if the mechanisms of failure and load redistribution are to be accurately reproduced in
reserve strength analyses. In addition, the benefits of hindcasting to benchmark both
deterministic and probabilistic evaluations of reserve strength are confirmed and will be
further examined in the review of Puskar et al. 1994.
1470
Broadside
1355
II
50 year storm I 1220 1615 I 1650 I
year storm 1335 1845 2060
I I I
The reserve strength ratios, in the range 1.1. to 1.4, are presented with respect to present
day year criteria. These values low in absolute terms but may be considered to be
substantial when the original 25 year design premise is considered. The basis of comparing
must be re-emphasised here, as definitions are often used with respect to the original
design values (see Section 2).
Corresponding with the findings of et the authors note that waves impacting
the deck were found to contribute as much as 30% of the total platform loading under the
100 year storm.
The paper demonstrates the versatility of reserve strength technology to encompass a wide
range of offshore structural forms and evaluate structural safety. Calculations of the
probability of failure were demonstrated to be in line with the proposed requirements of
API RP 2A (draft 1993) for fitness for purpose.
Structure RSR
SP62-B
SNS Inde-K
This table demonstrateshow first member failure may be synonymous with system collapse
whilst alternatively there may be significant potential for load redistribution
within the framing. The range of contributing factors means that reserve strength ratios
with respect to the original design basis vary from 1.91 to 3.78. Were a different
modelling or analysis strategy adopted (eg. in relation to material properties) there would
be a direct influence on the level of RSR calculated.
An important factor is the dependence of RSR on the original design basis. This was
highlighted in the original discussion (Section 6) and is illustrated in the paper for the Tern
structure. Different dead and environmental load factors in LRFD practice contrast with
a constant margin of safety in WSD. The authors compare the situations for loading in a
compression leg which governs the platform response for the diagonal wave attack
direction. It is demonstrated that RSR is related to the proportions of dead (P,) and
environmental (P,) loads for WSD and LRFD by the expressions.
For dead to environmental load ratios in the range 1.6 to 0.4, these expressions
imply greater RSRs for an LRFD design by 11 to 17 over WSD levels.
In examining Table A2.6, the RSRs with respect to the original base shear design loads
are instructive but meaningful interpretation depends on an evaluation of failure
probabilities or some relation with the associated return period.
Van de (1994) focuses on the Gulf of Mexico SP62-B, Southern North Sea Inde-K
and Northern North Sea Tern structures for this comparison. The best estimate year
loads are denoted L,, and the ratio of the ultimate strength compared with 100 year loads
is therefore given by RSR
Table A2.7 presents the values of for each of the platform locations together with
governing RSRs from Table A2.6. These lead to a much wider spread in reserve strength
ratios evaluated against present day 100 year criteria. Figure A2.4 presents these values
against the long term environmental statistics which demonstrate a lesser increase in loading
with the limited fetch of the Southern North Sea compared with the hurricane environment
in the Gulf of Mexico. These measures of reserve strength therefore imply significantly
different probabilities of failure as indicated in Table A2.7. Conversely for a consistent
safety margin (probability of failure) the required RSR in the Southern North Sea is less
than for the Gulf of Mexico (Efthymiou et al, 1994)
Table A2.7
Relation between reserve strength, probability of failure and environment
Together these papers demonstrate by example the care that needs to be taken in
interpreting reserve strength ratios. The values may be dependent on the underlying design
code and basic assumptions in the ultimate strength analysis. Furthermore the interpretation
of RSR in terms of safety depends on the environmental and long term, loading statistics.
As noted in the main report and this Appendix, the draft Section 17.0 to API RP 2A-WSD
admitted the use of ultimate strength analysis as part of the assessment of existing
platforms. The draft was subject to wide industry consultation and 1993) and
validation et al, 1995) and following four revisions was prepared as an API white
draft for comment (1995). However in April 1995 it was decided that the provisions could
be submitted for postal ballot without a formal draft issue and it is therefore anticipated that
the provisions of Section 17.0 (Section R in LRFD) will become part of API RP 2A in late
1995. Indeed, the Minerals Management Service have permitted its use in advance of
formal API balloting (Dyhrkopp, 1995).
Within Section 17.0 ultimate strength analysis forms just one part of the assessment
process. If fitness for purpose can be demonstrated on the basis of a design level analysis
or an elastic analysis with sources of conservatism removed, nonlinear analysis need not
be performed. However, the sources of reserve strength within the system revealed by
rigorous nonlinear analysis may be exploited in the assessment of existing platforms. The
approach is a significant departure from design practice embodied in API RP 2A to date
and the many difficult issues raised have been documented in a series of OTC papers by
the Task Group, 92-5, which drafted Section 17.0. These papers et al; Kallaby
et al; Petrauskas et al; and Turner et al; Kallaby and 1994) and
the related BOSS paper et al, 1994) present essential background to the user of
Section 17.0.
Of relevance to this review of reserve strength technology are the criteria for evaluating and
accepting ultimate strength assessments. They are particularly important because of the
definition of RSR employed and the specific loading regime adopted. The background to
these aspects of Section 17.0 is presented here, together with the results of background
analyses providing insight to representative jacket responses. The basis of acceptance
criteria is given and procedures for determining appropriate criteria outside US waters are
documented.
Definitions et al, 1994)
Reserve Strength Ratio in API RP 2A Section 17.0 is intended to provide a measure of
platform reliability in a given environmental region. It is therefore defined in relation to
present day design loads, ie:
Ultimate Strength Load
Eqn (A3.1)
API RP 2A 20th Edition year design load
In many earlier references covered in this review (see also Section 2), RSR was related to
the original design load and this important distinction should be noted.
Underlying Section 17.0 is the recognition that RSR comprises two distinct elements:
1. The change in design resistance and loads from the original design to the 20th Edition.
2. The system strength reserve beyond the maximum design elastic utilisation to the
nonlinear ultimate capacity.
These elements are quantified as follows:
Load Reduction Factor (LRF)
lobal load giving unity check of l to RP 2A 20th Edition
API RP 20th Edition 100 year design load
Note: changes in design criteria need not result in a reduction, although this is the
case in US waters for which Section 17.0 was initially developed.
Ultimate to Linear ratio (ULR)
Ultimate strength load Eqn (A3.3)
global load giving unity check of 1.0
As identified in discussion of work by van de et al (1993, 1994) in Section A2
above, the sources of system reserve (ie. ULR in Section 17.0 parlance) can be clearly
defined. The factors and ranges of values embodied in Section 17.0 are as follows:
code safety factors ( l .2-1.3)
mean to nominal yield strengths (1.2 for mild steel)
system redundancy (1.O-1.3)
designer's prerogatives (unspecified)
strength provided for other criteria (unspecified).
These lead to minimum values for ULR in the range 1.6 to 1.8 and tie in with the
findings of et al (1993). However, in other regions the contribution from the
different factors may vary, demanding that ULR be carefully defined at an appropriate
level.
It is then the combination of LRF and ULR (Equations A3.1 and A3.3) which leads to the
calculation of RSR:
RSR = LRF ULR Eqn (A3.4)
This procedure is clearly distinct from the definition of an RSR value per se.
Level Description
L3 low consequence
Platforms in the Gulf of Mexico are evacuated when there is a threat of hurricanes, so
environmental impact drives the most critical category. Level 1. Based on experience, all
47 platforms to have collapsed in hurricane conditions were designed to 25 year criteria
whereas structures designed to the 9th Edition of RP where the 100 year hurricane
criterion was introduced, and beyond had survived. It was decided that acceptance of
existing installations should be measured against the satisfactory performance of these
platforms. In going back from the new environmental loading recipe in the 20th Edition
of API RP 2A to the 9th-19th Edition loads, a Load Reduction Factor of 0.65 resulted.
This and the minimum ULR index of 1.6-1.8 presented above, give a target
RSR in the range 1.O-1.2 (eg. 1.2). Based on failure survival data from
Hurricane (Puskar et al, 1994) the Task Group adopted a value of 1.2. Thus the
derivation of RSR is largely experienced based.
For other classification levels the same principles of setting acceptance criteria with respect
to present day (20th Edition) design practice and working through to an RSR (based on an
appropriate ULR for the platforms under consideration) were followed.
The RSR criteria were not based on calculated target probabilities of failure, although
subsequent evaluations confirmed that the reliability was consistent with expectations.
A final consideration in the derivation of Gulf of Mexico criteria was the large number
(some 3,800) of platforms in a small area. (Indeed the justification for the criteria was the
significant experience base on which to draw). In applying the assessment criteria it made
sense to translate the for different classification levels to environmental parameters
corresponding to a with respect to the target RSR. In less densely populated areas
assessment criteria are more appropriately specified as target values such as RSR.
Summary
In relation to reserve strength, the subject of this review, it is clear that Section 17.0
provides significant acceptance of the technology. However the definition of RSR in
Section 17.0 differs from that generally used in the literature to date. The inherent reserve
strength of a platform beyond its original design is more closely reflected by the ULR
(Ultimate to Linear Ratio) in Section 17.0.
Many of the key areas identified in Section 6 of this report have been addressed to some
extent by Section 17.0, for example:
Application of realistic loading profile
Importance of realistic joint and foundation characteristics
Dependence of meaningful reserve strength evaluations on environmental and
design code basis
Uncertainty and importance of wave in deck load calculations.
However the user of Section 17.0 must recognise the US basis of the original provisions.
For application outside US waters it is essential that sources of system reserve in
representative structures are evaluated to give appropriate In addition the
acceptance criteria relative to new design (the LRF) must be revisited. Finally a rigorous
evaluation of risks with the consistency and methodology put forward by De (1995).
drawing on regional environmental criteria and base shear statistics is an essential pre-
requisite to the international application of 'Section 17.0' guidelines (Billington and Bolt,
Two significant benchmark studies have been conducted since the recommendation was
made in Section 7 of this review. The first has been undertaken by the UK Health and
Safety Executive (HSE, 1993 and 1994; Nichols et al, 1994) against the SCI frame test data
(Section 3). The second, commissioned by the US Minerals Management Service (MMS),
combined the trial applicationof Section 17.0 with a benchmark analysis of a specified Gulf
of Mexico structure. The scope and findings of these studies are reviewed in this section.
In preparing this review (see also Billington et 1993; Bolt et al, 1993) the need and
potential value of providing to industry the opportunity to benchmark available techniques
for reserve strength analysis against the SCI Frames Project test data (Bolt, 1994) was
identified. HSE managed the benchmarking exercise with external support provided only
by the Marine Technology Support Unit (MaTSU). Together HSE and MaTSU assimilated
and interpreted the results, presenting their findings in late 1994 (Nichols et al).
Eleven organisations participated in the exercise. Each was provided with a data package
giving the frame geometries, loading and support conditions, section sizes and
measured material yield stresses. Three base cases corresponded to Frames I, and
(see Figure 3.12 to 3.23). A fourth case specificed initial out-of-straightnessof braces and
locked-in stresses which had been identified as influencing the actual test results. The
structural responses of the various test cases (see Section 3) were:
Test Case 1 - Compression brace buckling
Test Case 2 - X joint failure followed by compression brace buckling with
significant portal action in frame legs
Test Cases 3 4 - Sequence of top bay compression brace buckling, load
redistribution and compression brace buckling in bottom bay.
In assimilating the results, grouped the predictions into three response types:
Type 1 - Poor agreement missing key features of the physical response.
Type 2 - Moderate agreement - key response characteristic captured but inaccurate
replication of load redistribution.
Type 3 - Good agreement - key response characteristics and load redistribution
captured.
With this categorisation the graphs in Figure A3.1 bracket the results obtained, noting that
each line may represent a set of similar predictions. At first sight the predictions appear
to be inconsistent, however it is important to investigate and understand the sources of
difference. This cycle in learning and benefitting from the benchmark exercise has not yet
taken place and the following comments are made by the authors of this review based solely
on the published results in Figure A3.1 and without sight of the individual analysis results.
Nevertheless the following observations are made to help explain the sources of
inconsistency.
Test Case
As noted in Section 3, the yield capacity of the chord and buckling capacity of the
compression brace were similar in the test frames making the response characteristic very
sensitive to effective length selections. Information on the post-buckling capacity of
members is sparse (eg. Pike and Grenda, and is influenced by many factors
including member slenderness, local buckling, out-of-plane deformations and material
modelling assumptions. It would appear from the results that the prediction of
buckling behaviour was a principal source of discrepancy.
However, it should be noted that the frame simplicity particularly highlights such
discrepancies. The frame test data therefore provide a valuable opportunity to explore these
influences and improve modelling practices. Whilst the load shedding characteristics are
less apparent in the overall response of a jacket structure, accurate modelling remains
important to ensure that loads are redistributed giving the correct sequence of subsequent
component failures.
Test Case 2
Response Types 1 and 2 merely indicate the absence of joint modelling in the analyses.
Code checks to API (1993) or HSE (1990) guidelines would have indicated the high
utilisation of a primary X joint. The conclusion is that the potential for tubular joint
failures must be considered in preparing structural analyses, given the potential significance
it may have on the global response characteristics.
By contrast, the achievement in obtaining results of Type 3 should be noted. The X joint
in Frame compressed to the extent that both braces came into contact across the joint as
shown in Figure 3.17. This enabled the pick up in load and subsequent buckling indicated
in the experimental and Type 3 responses. Isolated test data only reveal the initial yield
characteristic of the X joint but the benchmarking illustrates the need for careful monitoring
of component deformations in predicting the response of framed structures.
Test Case 3
Discrepancies between the results again reflect the different modelling of post-buckling
member capacities. Where the post-buckling capacity remains high insufficient loads are
transmitted into the bottom bay to precipitate the sequence of buckling.
The different load levels and deformations associated with the buckling in the Type 3
predictions compared with experiment, are attributed to the locked-in stresses and initial
imperfections in the test frame.
On the basis of this discussion of the responses in Figure A3.1, the differences between test
results and the predictive capability of software and users can begin to be explained. It
appears there may be only a few key features underlying the divergence in results. Firm
conclusions on the industry capability must await detailed investigation and quantification
of these effects by the analysts in open forum.
A5 CONCLUSIONS
The discussion and conclusions presented in the main review (Sections 6 and 7) remain
valid and may be read in conjunction with this Appendix. A number of additional points
may be made in relation to the developing application of the reserve strength technology.
Practitioners in operating companies and development organisations alike recognise the
complexities in performing ultimate strength analysis, as witnessed by the continuing
emphasis on simplified approaches to provide insight to the probable mechanisms of
system reserve et al, 1994; et al, 1994; Bea, 1995). It is therefore
important that industry knowledge and experience continues to develop and becomes
embodied in best practice guidelines such that the appropriate expertise leads to
meaningful results.
Benchmarking against test data (HSE, 1994) and comparative analyses et al,
1995) provide an important vehicle to advance understanding of the technology but
depend on openness and cooperation between organisations for any benefits to be
derived.
Discrepancies between ultimate strength predictions should not be used to discredit the
technologies, but should underline the skill and capabilities required of software,
analysts and engineers both in interpreting results and in commissioning ultimate
strength evaluations. Furthmore due allowance should be made in the setting and
implementation of ultimate strength criteria to account for the potential variability due
to software differences and user selections or interpretations. Best practices should be
identified and carried forward and inadequate methods rejected on the basis of industry
experience and consensus.
Whilst Section 17.0 provides a framework for ultimate strength analysis in the process
of platform assessment, it is essential that the regional dependence, in terms of both
reserve strength and environmental loading, be recognised and efforts made to develop
criteria for the safe application of the technology worldwide.
Section 17.0 quantifies ultimate system responses in terms of a single deterministic
capacity measure, the Reserve Strength Ratio. The main review questioned the
sufficiency of RSR given, for example, the potential importance of deflection
serviceability limit states and sensitivity of ultimate strength predictions depending on
failure mode. With the progression of Section 17.0 to the basis of an standard this
point should be re-emphasised and some account taken of the inadequacy of RSR as a
lone assessment criterion.
Reserve strength predictions for offshore structures depend on accurate data for
member, joint, foundation and fluid load parameters not only at the point of failure but
into the post-ultimate regime. Efforts must continue to obtain data representative of
the large deflection conditions within the constraints of the system and meanwhile
rational account must be taken of the modelling uncertainties.
Figure A2.4
The relation between reserve strength, environment and probability of failure
.
, -- Experimental
1
I
-2
0
0 0.05 0.1 0.15 0.2 0.25 03
Lateral displacement (m)
0 I
0 0.05 0.1 0.15 0.2 0.25 0.3
Lateral displacement (m)
Figure A3.1
Classification of HSE benchmarking results against Frames Project data et
I
Figure A3.2
Benchmark jacket structure et al, 1995)
Figure A3.3
Load displacement predictions for the benchmark platform et 1995)
Printed and published by the Health and Safety Executive
C2 6/96