Académique Documents
Professionnel Documents
Culture Documents
Al Conle
University of Windsor, Windsor, ON, Canada
ABSTRACT
The in-service fatigue loading of risers may include
contributions from first and second order vessel motions, wave
loadings, vortex induced vibration (VIV), vortex induced motion
(VIM), pressure variation and slugging. All of these loadings are
variable amplitude random loadings where the sequence of cycles can
change the fatigue outcome. The loading sequence effects can be
addressed in analysis through cycle counting methods and through use
of the most appropriate methods of damage accumulation. High loads
can be responsible for either fatigue crack growth retardation or
acceleration depending up loading sequence. The Rainflow Counting
method takes into account the whole loading algorithm and assumes the
structure memory sustains all through the loading. The Simple Range
Count method counts cycles by every single reversals and does not take
into account the load sequence or material memory. Fatigue damage
accumulation can be accomplished using a linear model like PalmgrenMiner or a nonlinear model. Fatigue life can be estimated using crack
growth methods, S-N or strain life methods. For each method some
tools are available that can account for sequence effects. The fatigue
analysis of the riser system is illustrated step by step. This study
particularly looks at how different cycle counting methods affect the
fatigue damage estimate obtained with the S-N approach, strain-life
approach and fatigue crack growth approach. In the study Rainflow and
Simple Range cycle counting methods are applied to variable amplitude
loads typical of deepwater riser systems.
INTRODUCTION
Risers experience variable amplitude fatigue loading caused by first
and second order vessel motions, wave loadings, vortex induced
vibration (VIV), vortex inducted motion (VIM), pressure variation and
slugging. Developing a realistic representation of service load is a key
step to successful fatigue analysis and design of deepwater riser
systems.
My
Mz
I
R
histories.
Table 1 Rainflow Counting Result
Cycle
A-D-I
B-C-B
E-H-E
F-G-F
Range
39
9.5
28
5
No. of Cycles
0.5
0.5
0.5
0.5
0.5
0.5
0.5
0.5
(2)
Where,
N
a,m
k
tref
(3)
Where,
max
a
E
'f
'
f
c
Nf
In the crack growth approach for riser fatigue assessment, the analysis
is conducted based on the Paris Law approach, as defined by the
Equation (4). Just as for the SN method it can be implemented with
multiple slopes and take mean stress into account:
(4)
da /dN = A (K)m
Where:
da/dN
is the fatigue crack growth rate
K
is the stress intensity range
A,m
are the parameters defining the single or multi stage crack
growth relationship
In both S-N and strain-life approaches, after fatigue life N for the stress
range / strain range is calculated for each bin or count, the damage for
each bin can be calculated given the number of cycles in that bin
computed by cycle counting. The cumulative damage can be calculated
by the Palmgren-Miner rule (Palmgren, 1924, Miner, 1945) or a
nonlinear model (Yin and Fatemi, 2010; Yang and Fatemi, 1998). The
Palmgren-Miner rule is expressed in the following equation:
(5)
Although there is experimental and theoretical evidence that indicates
that linear damage accumulation is not always accurate, PalmgrenMiner rule is still the most commonly used approach in S-N and strainlife fatigue damage calculation not only due to its simplicity but also
because all other nonlinear models require testing data to validate the
input parameters. The complexity of each service load makes it hard to
obtain one single nonlinear model that fits all service histories.
The fatigue crack growth approach is usually adopted in riser design --via engineering critical assessment (ECA). For a riser system already in
service, the fatigue crack growth approach is also selected for fitness
for service (FFS) assessment. In general a fatigue crack growth analysis
with variable amplitude loading depends on the load sequence
especially when isolated overloads occur in the loading history. A
tensile overload may induce compressive residual stress fields that tend
to temporarily retard or arrest crack growth. A compressive overload,
on the other hand, may leave tensile residual stress fields that tend to
accelerate crack growth. In ECA or FFS of risers, the load history is
usually generated from the statistics of the expected service life and the
loading is randomized by assuming that a characteristic histogram or
load block is repeated. The block load histogram is obtained from
global analysis and cycle counting. However when severe overload
occurs (this can happen when a hydro-test is performed) or when a load
amplitudes change magnitude in predominantly one direction (high
blocks followed by low blocks or low blocks followed by high blocks),
fatigue crack growth models incorporating load sequence effects need
to be incorporated. When load a history is highly random, the
assumption is usually made that the hi-lo and lo-hi effects cancel out
each other This ability of the FCG method to account for load sequence
effects has its advantages compared to S-N and strain-life methods.
CASE STUDY
As an example of how the cycle counting method affects the predicted
fatigue life, an analysis is conducted with a deepwater steel catenary
riser (SCR) in the Gulf of Mexico (GOM). The loading is derived from
a 3 hour time domain simulation of a 100 year hurricane. The SCR
configuration is shown in Figure 5. Two locations along the SCR are
identified as fatigue critical locations: 1. The first (top most) weld near
hang off; 2. The touch down point (TDP).
Figure 6 Stress Time Traces for First Weld and TDP under 100 Year
Hurricane
Value
Steel
X70
320
38
482
565
207,000
0.3
The three hour time traces are assumed to repeat in fatigue analysis.
Stress histograms obtained with both Rainflow counting and Simple
Range counting methods are shown in Figure 7 for TDP and Figure 8
for first weld. In smaller stress ranges the difference between the
Rainflow counting and Simple Range counting results is not significant
for these two stress time traces. The difference starts to show in larger
stress ranges which contribute higher fatigue damage. It can be seen
from both Figure 7 and Figure 8 that Rainflow counting results in more
large stress range cycles than Simple Range counting, which would
lead one to conclude that Simple Range counting could result in underconservative results in fatigue damage calculation.
tref
(mm)
SCF
3.0
22.0
0.25
1.1
Figure 10 Fatigue Crack Growth Curves for the First Weld (Rainflow
Counting vs. Simple Range Counting)
CONCLUSION
Load sequence effects sometimes play an important role in fatigue
estimation in variable amplitude loading. Before conducting the fatigue
assessment for the riser system it is recommended to screen the load
histories either from the time-domain simulation or from the real-time
monitoring data to check if one-sided overloads are expected to
produce sequence effects.
Different cycle counting methods can result in very different load
histograms for fatigue assessment. Caution should be taken to select the
appropriate cycle counting method. Rainflow counting method is the
most widely used cycle count method and is recommended to be used,
especially when no experimental testing data are available to justify the
material memory has been lost during the loading history.
The S-N approach is most widely used in fatigue assessment of riser
system. When local stress concentration is expected (for example in
mechanical connectors), the strain-life approach is recommended. Both
S-N and strain-life approaches are used to predict life to crack
initiation. However when crack growth life is a significant part of the
total fatigue life, fatigue crack growth approach should be used.
In all three fatigue assessment approaches: S-N, strain-life, and fatigue
crack growth, Simple Range counted histogram results in 30% to 200%
higher lives compared to Rainflow counted histograms. Using Simple
Range counting could result in non-conservative riser design.
REFERENCES
Palmgren A.G.: Die Lebensdauer von Kugellagern (Life Length of Roller
Bearings. In German). Zeitschrift des Vereines Deutscher Ingenieure
(VDI Zeitschrift), ISSN 0341-7258, Vol 68, No 14, April 1924, pp
339341.
Miner A.M.: Cumulative Damage in Fatigue, J. of Appl. Mech. Trans
ASME, 12, 159-164, 1945.
ASTM Standard E 1049 - 85 (2005). Standard Practices for Cycle
Counting in Fatigue Analysis.
Matsuishi, M. and T. Endo, "Fatigue of Metals Subjected to Varying
Stress," presented at Japan. Soc. of Mech. Engrg., Fukuoka, Japan,
March 1968.
BS5400: Part 10, 1980, "Cycle counting by the reservoir method,"
Appendix B, pg. 9/22.
Downing, S.D. and D. F. Socie, "Simple rainflow algorithm," Int. J. of
Fatigue, V4, N1, 31-40, 1982.
Conle, A., T.R.Oxland, T.H.Topper, "Computer-Based Prediction of
Cyclic Deformation and Fatigue Behavior," Low Cycle Fatigue ASTM
STP 942, 1988, pp.1218-1236
OTO 199 058, Fatigue Design Curves for Welded Joints in Air &
Seawater Under Variable Amplitude Loading. HSE, July 2000.
Smith K.N., Watson P. and Topper T.H., A Stress-Strain Function for
the Fatigue of Metals, ASTM J. of Materials, V5 N4 Dec. 1970.
SAE Fatigue Design Handbook, 3rd Edition, Editor R.C. Rice, Society