Professional Documents
Culture Documents
an unbiased predictor of the mean strength of X-grade pipe, thickness, the model was verified against ring tests and shown
with a standard deviation of the error of 3%. to be mean value correct with a standard deviation of less than
Although any general stress-strain law can be employed 3% [7]. The two-zone model has recently been shown (see
for burst strength predictions, fitting a Lüdwik power law companion paper by Klever and Stewart [5] ) to be an
curve to the material uni-axial true-stress/logarithmic-strain accurate predictive tool for the burst capacity of worn casing.
curve leads to the simplest approach (Stewart et al. [7]). This For plane strain, the two-zone burst capacity of a pipe with
curve, which is intended to fit the material data in the large thin section t * and thick section t can be expressed as:
strain region of interest, takes the form
p 2*z = f wl p * , (5)
n
σ = Cε (1)
where p * is the burst pressure of a pipe with uniform reduced
where σ is the true Cauchy stress and ε is the logarithmic wall thickness t * , given by
strain. The constants n and C are given by
t* t*
e
n p * = f 1 σ uts = p . (6)
n = ln(1 + ε ult ) ; C = σ uts , (2) R t
n
The wall loss enhancement factor f wl associated with the
with ε ult being the engineering strain corresponding to the two-zone model in Equation (5), is equal to or greater than
engineering ultimate stress, σ uts . unity and accounts for the increased capacity over the burst
Taking the average of the Tresca and von Mises burst pressure of a pipe with uniform thickness t * . It depends on
strength [7] results in the following analytical expression for
the strain hardening index, n, and the ratio of the arc-lengths
burst capacity of a pipe with uniform thickness t:
of the thin and thick zones. The case when the arc length of
t the thin zone tends to zero (as for an axially cracked pipe), is
p = f 1 σ uts (3)
R shown in Fig. 1 for P110/Q125 pipe. For deep cracks, f wl
where has a limiting value of 2n.
n +1 n +1
1 1
f1 = + , (4) Table 1 Burst test database on as-received pipe
2 3
Test OD t-min UTS Strain P-burst P-burst
p is the internal pressure differential at burst, and R is the at UTS test model
mean radius.
no. [in] [in] [ksi] [%] [ksi] [ksi] ratio
Burst data for OCTG pipe without flaws 1 7.064 0.352 107 11.5 11.3 11.3 0.997
As part of a test programme to establish the burst capacity of 2 3.530 0.254 109 7.7 17.3 17.4 1.008
worn pipe and pipe with cracks, two burst tests were carried 3 7.805 0.575 144 6.5 25.2 23.7 0.942
out by TNO on as-delivered pipe of grades P110 and Q125 to
provide base-line information. Recently, Paslay et al. [8] 4 7.068 0.524 131 6.1 22.1 21.8 0.985
tested a number of OCTG pipes. Their data on pipes without 5 7.076 0.408 105 9.3 13.4 13.1 0.976
flaws, together with the TNO data is shown in Table 1. In the 6 7.076 0.407 133 7.0 17.2 16.8 0.975
last column of this table, the ratio of the model prediction to
10 9.724 0.388 104 10.2 8.9 8.7 0.983
the test burst pressure is provided. The model given by
Eq. (3) (using t = tmin) is unbiased and has a standard deviation 11 9.929 0.531 102 10.2 11.8 11.7 0.988
of 2.7%. It is concluded that the analytical model of Eq. (3) 12 3.502 0.567 106 11.5 42.7 41.2 0.965
is a good predictive tool for the burst capacity of OCTG pipe 14 2.649 0.154 121 6.5 16.4 15.5 0.944
without flaws.
20 7.070 0.473 130 6.8 19.4 19.3 0.993
Plastic limit model for pipe with local wall loss 21 7.805 0.577 144 6.5 24.9 23.8 0.957
In his paper on burst, Klever [6] also developed a ‘two-zone’ 22 7.089 0.585 144 6.5 25.9 26.8 1.037
membrane tension model to predict the burst capacity of pipes TNO1 13.47 0.524 142 5.0 11.7 12.0 1.026
with smooth flaws. He assumed that the wall loss is long in
TNO2 9.921 0.606 155 6.0 20.7 21.0 1.014
the axial direction and that the pipe can be represented by a
plane model represented by two arcs (zones), each having mean 1.00
different wall thickness. Uniform straining through the std. dev. 0.027
thickness was assumed in each zone. For the case where the
rc-length of the thinner zone is large compared to its
SPE 48830 ACCOUNTING FOR FLAWS IN THE BURST STRENGTH OF OCTG 3
1.050 ε / ε o = α (σ / σ o ) 1/ n , (7)
capacity of pipe with thickness t*
1.040
π
J p = ασ o ε o a (1 − a / t ) ( P / Po ) 1+1/ n (8)
1.030 n
Q125 where a is the crack size, P is the applied load, and Po is the
1.020 P110 load at which the remaining ligament yields based on a yield
stress of σ o . For the case of cracked pipes, P ≡ p is the
1.010 internal pressure (differential), and the pressure po , is taken
to be that predicted from von Mises yield criterion such that
2 t−a
1.000
po = σo. (9)
0.00 0.05 0.10 0.15 0.20 0.25 0.30 3 R
w all loss/original thickness
By comparing Equations (1) and (7), it is straightforward to
show that
Fig. 1 Factor f wl as a function of wall loss
n
α≡
eε o
(σ o / σ uts ) 1/ n . (10)
When the length of the thinner zone is of the order of the
wall thickness, such as would be the case with an axial part- This leads to the following expression for J p for the cracked
through crack, the assumption of uniform through-thickness
pipe:
straining leads to a two-zone limit load that is somewhat less
than that predicted by finite element models that account 1+1/ n
properly for shear banding. Therefore in the absence of a
1 p
fracture mechanism, where limit load behaviour governs, the J p = a (1 − a / t )π nσ uts . (11)
two-zone model provides a lower bound estimate to the burst e 2 t −a
σ uts
strength of an axially cracked pipe. 3 R
We will make use of this relationship later for the assessment
3. J-INTEGRALS FOR PIPES WITH
of burst tests on cracked pipes to determine whether the burst
CRACK-LIKE FLAWS
capacity is limit-load or fracture controlled.
Real pipes will always have small flaws, the size of which will
depend on the manufacturing process and the quality of
Comparison with F.E. results
inspection. For pipes with cracks, there are two competing
failure mechanisms. One is the plastic limit-load behaviour To assess the validity of using the Hutchinson infinitesimal
described above. The other is fracture, which we now solution for the plane strain crack for a pipe with an axial part-
consider. through external flaw, large strain J-integral F.E. calculations
The J-integral method (see Rice [9]) provides a single were undertaken for P110 and Q125 casing. These analyses
parameter for estimating the onset of ductile fracture in pipes were performed on behalf of Shell by MARC Research
containing flaws and we look to this approach to augment the Europe. The geometry and material properties of the casing
models developed above. The J-integral approach is not studied matched that of pipes tested by Mitsui-Babcock with
strictly valid when the remaining ligament strains in the artificial flaws, details of which are given in Table 2a below.
plastic range (loss of J-dominance); however in this case, the A von Mises yield criterion was employed in these analyses.
toughness can be considerably enhanced owing to loss of The results are shown in Figs. 2a and 2b.
constraint ahead of the crack tip, as discussed by Odow’d and It is clear that the Hutchinson formula performs very well
Shih [10, 11]. Consequently, the application of the J-integral for both material grades. The normalising pressure in these
assuming full constraint at the crack tip is conservative. figures is the von Mises two-zone limit load which is defined
as
The J-integral solution for a crack in plane strain t*
p 2*z _ vm = f vm f wl σ uts (12)
Hutchinson [12] has developed a small strain solution for the R
plastic component of J for a plane strain crack in an infinite
where t * is the remaining ligament thickness and
body. Ignoring the elastic contribution (which is small when
the ligament strains in the plastic range), and assuming a n n +1
1 2
power law fit to the material uni-axial stress/strain curve of f vm = . (13)
the form 2 3
4 G. STEWART, F.J. KLEVER SPE 48330
p FE = f FE p 2*z (14)
800
where p FE is the limit load calculated by F.E. analysis and
700 f FE is the increase over the two-zone model that results from
FE three dimensional effects.
600 Analytical Further details on F.E. comparisons with the two-zone
model can be found in Klever and Stewart [5].
500
4. BURST TESTS FOR OCTG PIPE WITH
J [KPa.m]
500
in Table 2a, while the test results are compared with limit load
400 predictions in Table 2b. In Table 2a, tmin is the thickness at
the location of the flaw before the flaw was inserted while t*
300 is the ligament thickness at the flaw location given by
t* = tmin - a.
200 In Table 2b, P*_2z is the limit capacity calculated from
100 Eq. (5). The F.E. correction factor corresponds to f FE in
Eq. (14). It is observed that the two-zone model underpredicts
0 the burst pressure recorded in the tests. When the prediction
0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 is adjusted to account for the limit load predictions based on
P/P*_2z the F.E. analysis, the answers are within 4% of the recorded
test burst pressures. This is similar to the agreement found
Fig. 2b F.E. J-Integral for Q125 pipe with external between limit load models and tests on regular pipes.
crack compared to Hutchinson plane strain The burst pressures for these cracked pipe tests show only
solution a small reduction over those for the defect free pipes tested by
TNO given in Table 1 (for which the P110 pipe burst at
As mentioned previously, the two-zone model 80.6 MPa and the Q125 pipe at 142.9 Mpa). This confirms
underestimates the limit load, in this case by 6% for both that failure is by a plastic limit load phenomenon.
pipes, as may be seen from Figs. 2a and 2b. This can be
expressed as follows:
SPE 48830 ACCOUNTING FOR FLAWS IN THE BURST STRENGTH OF OCTG 5
J {KPa.m]
300
200
Table 2b Burst capacity of test pipes and
limit loads from models 100
100
is predicted that the burst pressure would have been 3% lower.
80
The picture for the Q125 test is somewhat different. In
60 this case, the fracture analysis (Fig. 5a) predicts that the pipe
40 with the sparc-eroded defect should burst at 93% of the
20 P110 Q125 maximum pressure recorded during the test. Indeed, to
sustain the test pressure, the resistance curve would need to be
0
scaled by about a factor of 3. There may be several reasons
-40 -30 -20 -10 0 10 20
why this increase in resistance can be justified. The most
Tem perature (C) credible is that for the edge cracked pipe, the constraint in the
Fig. 3 Charpy impact transition curve fracture process zone ahead of the crack is substantially less
than that in the small scale 3-point bend J-R specimen. This
In the next section, we investigate the crack driving force loss in constraint reduces the magnitude of the mean tri-axial
and compare this to the resistance measured from fracture test stress which leads to an increase in apparent fracture
specimens to determine whether the absence of fracture is toughness during stable tearing. This concept of two
indeed predicted and to assess the consequences of having parameter fracture criterion has received considerable
fatigue sharpened cracks as opposed to the sparc-eroded attention recently and is still evolving. Notable publications
notches used in the test pipes. in this area are those by Odow’d & Shih [10, 11], MacLennan
& Hancock [13] and Tvergaard & Hutchinson [14], to name
Fracture assessment of test pipes but a few. Increases in toughness (Jmat) by a factor of 4 have
Figure 4 shows the results from a ductile tearing assessment been reported in the literature [13].
for the P110 pipe where the applied J values have been As with the P110 material, the fracture resistance of the
calculated according to Eq. (11) and the (lowest of the pre-fatigued Q125 specimens is lower than the sparc-eroded
recorded) resistance curves for both sparc-eroded and pre- specimens (compare Figs. 5a & 5b).
fatigue cracked specimens are shown. It is observed that the
sparc-eroded specimen (R_sparc) has a significantly larger
resistance curve than the pre-fatigued crack specimen
(R_fatigue).
6 G. STEWART, F.J. KLEVER SPE 48330
1600 assumed that the crack is sharp (which is not always the case
J_P-test for manufacturing defects), and hence the resistance curves
1400 J_0.93*P-test
with the pre-fatgued crack are employed.
R_sparc
1200 R_sparc*3
The procedure followed was to take an initial crack size a,
and to increase the pressure until the applied J curve was
1000
tangential to the resistance curve. This pressure is denoted
J [KPa.m]
800 the fracture instability pressure. For all initial crack sizes,
this gave a crack extension of 0.2 mm for the P110 pipe and
600
0.3 mm for the Q125 pipe. Using the two zone model of
400 Eq. (5), the limit load was then calculated for the remaining
ligament thickness including crack extension. The lower of
200
the limit-load and the fracture instability pressure is taken to
0
be the burst capacity. In those cases where the limit-load
2 2.2 2.4 2.6 2.8 3 3.2 governs, this procedure yields a conservative two-zone
a [m m ] capacity prediction since the crack extension at the limit
pressure may be less than that associated with the fracture
Fig. 5a Tearing assessment of Q125 pipe
instability pressure. Note that the finite-element correction
with sparc-eroded defect
factor f FE is not used here as in general this factor will not be
1600
known. The procedure provides conservative estimates of
J_P-test
1400
J_0.91*P-test
failure capacity for both limit behaviour and ductile fracture.
J_0.97*P-test The results of this assessment are plotted in (Fig. 6). The
1200
R_f atigue horizontal axis is the initial crack depth to wall thickness ratio,
1000
R-f atigue*3
while the ratio on the vertical axis is the applied pressure to
the burst pressure of a uniform pipe with thickness t * = t - a,
J [KPa.m]
800
where a is the initial crack size. It may be seen that for the
600 P110/Q125 pipe samples considered, the pressure p *
corresponding to the limit load of a pipe of remaining
400
ligament thickness t * provides a good estimate of the burst
200 capacity envelop calculated from the fracture and limit-load
0
models. Note that the ratio of the limit-load to p * is less
2 2.2 2.4 2.6 2.8 3 3.2 than that shown in Fig. 1 because in this case it includes crack
a [m m ] extension. It is interesting to note that the fracture instability
curves for the two materials fall more or less on top of each
Fig. 5b Tearing assessment of Q125 pipe other.
with pre-fatigued notch
1.05
plastic. 0.97
0.96
5. FAILURE DIAGRAM FOR P110/Q125 PIPES 0.000 0.050 0.100 0.150 0.200 0.250 0.300
1
Table 3 Parameter in exponential distribution
0.9 Reject/repair frequency (ν) Inspection λ
0.8 quality
Probability of detection
(19) 4
1
λe − λη
∫ (1 − α e
−β η
= ) dη
η min
1 − e −λ 3
2
Equation (19) can be explicitly integrated, although it is
more convenient to perform numerical integration. If the 1
repair rate is known (or estimated) the parameter λ in the
initial flaw distribution can be established. Table 3 provides 0
0 0.03 0.05 0.08 0.1 0.13 0.15 0.17 0.2 0.22 0.25
values of λ for various repair rates and inspection qualities. Crack-size/wall-thickness
initial distrib.
updated distrib.
Technology Workshop on Risk Based Design of Well Casing Journal of App. Mech., Vol. 35, pp 379-386, 1968.
and Tubing, paper 48329, The Woodlands, Texas, USA, May, 10. Odow’d, P.N., and Shih, C.F., “Family of crack-tip fields
1998. characterised by a triaxiality parameter - I. Structure of fields”,
6. Klever, F.J., “Burst strength of corroded pipe: flow stress J. Mech. Phys. Solids, Vol 39, No. 8, pp 989-1015, 1991.
revisited”, Offshore Technology Conference (OTC), Houston, 11. Odow’d, P.N., and Shih, C.F., “Family of crack-tip fields
Texas, 1992. characterised by a triaxiality parameter - II. Fracture
7. Stewart, G., Klever, F. J., and Ritchie, D., “An analytical model applications”, J. Mech. Phys. Solids, Vol. 40, No. 5,
to predict the burst capacity of pipelines”, Offshore Mechanics pp 939-963, 1992.
and Arctic Engineering (OMAE) conference, Houston Texas, 12. Hutchinson, J.W., “Fundamentals of the phenomenological
1994. theory of nonlinear fracture mechanics”, J. App. Mech., 1042,
8. Paslay, P.R., Cernocky, E.P., and Wink, R., “Burst pressure Vol. 50, December, 1983.
prediction of thin-walled, ductile tubulars subjected to axial 13. MacLennan, I., and Hancock, J.W., “Constraint-based failure
load”, SPE Applied Technology Workshop on Risk Based assessment diagrams”, Int. J. Pressure Vess. & Piping, Vol. 64,
Design of Well Casing and Tubing, paper 48327, The pp 287-298, 1995.
Woodlands, Texas, USA, May, 1998. 14. Tvergaard, V., and Hutchinson, J.W., “Effect of T-stress on
9. Rice, J.R., “A path independent integral and the approximate mode I crack growth resistance in a ductile solid”, Int. J. Solids
analysis of strain concentration by notches and cracks”, ASME Structs., Vol. 31, No. 6, pp 823-833, 1994.