Professional Documents
Culture Documents
API/ISO TC67/SC5/WG2b: Modernisation of Api Bulletin 5C3 Api Prac Report: Determination of Stress-Strain Curve Shape For J/K55 Product
API/ISO TC67/SC5/WG2b: Modernisation of Api Bulletin 5C3 Api Prac Report: Determination of Stress-Strain Curve Shape For J/K55 Product
API/ISO TC67/SC5/WG2b: Modernisation of Api Bulletin 5C3 Api Prac Report: Determination of Stress-Strain Curve Shape For J/K55 Product
Issue 01
January 2007
Tel: 01224-351207
Fax: 01224-351100
E-mail: adrian_adams@nexeninc.com
API/ISO TC67/SC5/WG2b: Modernisation of API Bulletin 5C3 Page 2
API PRAC Report: Determination of Stress-Strain Curve Shape for J/K55 Product
TABLE OF CONTENTS
1.0 INTRODUCTION
2.0 SUMMARY
3.0 DATA
5.0 RESULTS
6.0 CONCLUSIONS
7.0 RECOMMENDATIONS
8.0 REFERENCES
This report has been prepared as part of service to the American Petroleum Institute, and any views of the
author expressed in it are not necessarily the views of his employer/Nexen. Neither the author nor his
employer/Nexen shall be liable for any reliance on its contents.
1.0 INTRODUCTION
Workgroup 2b of TC67/SC5 is currently revising API Bulletin 5C3 for recent developments in
pipe manufacture and resistance modelling. The work is shortly due for issue as TR10400 [1]. The
new collapse equations cover both quenched and tempered (Q&T) and non-Q&T pipe.
For Q&T pipe, the equations explicitly include all variables found to affect collapse strength
(outside diameter, wall thickness, yield stress, eccentricity, ovality, and residual stress).
For non-Q&T pipe, theoretical work [2, 3] suggests that collapse strength is also affected by
stress-strain curve shape (SSCS). Work therefore proceeded on two fronts; first, to check whether
the theoretical models agreed with collapse test results, and secondly to characterise the statistical
variation of SSCS for use in calibrating the design equations.
Both efforts were unsuccessful. Collapse test data showed no systematic variation in strength with
SSCS [4, 5], and a protracted measurement programme using production tensile test data gave far
too high a spread of calculated Young’s modulus to be credible [4-10].
It was suggested [5] that these apparent contradictions were due to inaccuracy of strain
measurement for tensile testing to production specifications [11-13]; and it was noted that tensile
test specifications for calculation of Young’s modulus [14] required much more accurate strain
measurement, which tended to bear out this view.
To confirm this hypothesis, it was decided to conduct a dual tensile test programme on sample
pairs cut from J/K55 product, with the first sample of each pair tested per API 5CT by the
contributing mill, and the second tested to high accuracy by the UK National Physical Laboratory
(NPL). The objectives of the study were to:
This report documents the study data, analysis methods, results and conclusions. The work was
undertaken in May 2004 - September 2005.
2.0 SUMMARY
A dual tensile test programme was undertaken, with the first sample of each pair tested by the mill
to API 5CT, and the second by NPL to ASTM E111. The results showed that:
• Tensile testing to production specifications gives accurate estimates of API yield stress
(±3%); inaccurate estimates of Young’s modulus (±30%); and highly inaccurate estimates
of stress-strain curve shape (±1000%).
• The errors in the latter are primarily due to lack of centralisation, and thus inaccurate
strain measurement for single-sided extensometry.
• High-specification tensile testing (to ASTM E111 or better) is required for accurate
determination of Young’s modulus and stress-strain curve shape.
Further testing is required in order to give data which can be used in design code calibration.
3.0 DATA
Twenty sets of samples of J/K55 cold rotary straightened product were tested, in two batches of
ten. The first batch was provided by an ERW mill, and the second by a seamless mill. Both held
current API licences. All samples were provided under confidentiality, and thus in the following,
each mill is represented by an alphanumeric code of their own choosing.
In each case, the sample pairs were cut from adjacent positions on the pipe. The first sample of
each pair was tensile tested by the respective mill, to the standard API 5CT/ASTM A370 test
specification [11, 12]. The tensile test data were sent in digital (Excel) format to the present
author, and Young’s modulus and stress-strain curve shape (SSCS) calculated.
The second samples of each pair were sent to the UK National Physical Laboratory (NPL) as
rough-cut blanks, and tensile tested to an NPL in-house specification exceeding that of ASTM
E111 [14]. The first batch of samples (from manufacturer CT03) was stress relieved, and the
second batch (from manufacturer XY01) was tested as-received. As before, the tensile test data
were sent in digital format to this author, and Young’s modulus and SSCS calculated.
Figures 1 and 2 give the stress-strain curves for the mill and NPL tensile tests. Also shown are the
fitted curves and numerical results; these are discussed in the sections following.
Young’s modulus was determined using Roebuck et al. 1994 [15]. This method was developed
for materials with short linear ranges, such as composites, but is currently also considered the
most accurate technique for metals1. Appendix B gives the MathCAD spreadsheet written to
implement the method. It was validated against independent calculations by NPL [16].
Needleman’s stress-strain curve [17] was used. In this formulation, the stress beyond the
proportional limit is given by:
1/ n
⎛ E ⎞
σ(ε ) = f p ⎜ n ε +1− n ⎟
⎜ fp ⎟
⎝ ⎠
where:
1
The author is grateful to Dr. Jerry Lord (UK National Physical Laboratory), for expert advice on
materials testing.
fp = proportional limit
ε = strain
E = Young’s modulus
n = hardening index
This approach results in n being wholly determined by a given combination of fp, E, fy and εy.
Appendix B gives the MathCAD spreadsheet written to implement the method.
The numerical results are given to the right of each stress-strain curve (Figures 1 and 2), and
summarised as Tables 1 and 2. Table 1 lists API yield stress for the manufacturer XY01 samples2.
Four columns are given, as follows:
• The first, headed “Prod. Mill” is for the mill tensile tests, with results calculated by the
mill.
• The second, headed “Prod. WG2B” is for the mill tensile tests, with results calculated by
the workgroup.
• The third, headed “NPL WG2B” is for the NPL tensile tests, with results calculated by the
workgroup.
• The last, headed “Difference” gives the percentage difference between the first and third
columns.
The maximum difference between the mill and NPL results is 2.8%.
Table 2 gives the Young’s modulus results. The production test values are very widely scattered,
with a mean of 30550 ksi and a COV3 of 0.0939. In contrast, the NPL results all fall within a very
narrow range, with a mean of 29910 ksi and a COV of 0.0048. The maximum difference between
the mill and NPL results is 23.9%.
This very much suggests that the wide range of earlier WG2B results [4-10] were due to
2
For manufacturer CT03, the NPL tests were for stress relieved samples, and therefore the results
can only be read for Young’s modulus.
3
Coefficient of variance (COV) is a dimensionless measure of dispersion, given by standard
deviation/mean.
measurement inaccuracy, rather than any variation in the material itself. How, therefore, does this
inaccuracy arise? Table 3 compares tensile test specifications for yield stress measurement (API
5CT, ASTM A370 and ASTM E8) with that for Young’s modulus measurement (ASTM E111).
The yield stress specifications allow single-sided extensometry (that is, only one strain gauge),
whereas the Young’s modulus specification requires double-sided extensometry (two gauges)4.
If single-sided extensometry is used, then any bending stress due to lack of sample centralisation
will be measured in addition to the true axial stress; and this will in turn cause an increase or
decrease in calculated Young’s modulus, according to whether the gauge is on the convex or
concave side.
With double-sided extensometry, sample bending still occurs, but its effect is negated by
averaging the readings of the two gauges. One should therefore expect double-sided extensometry
to give much more accurate results for quantities which rely closely on true axial strain, such as
Young’s modulus and stress-strain curve shape.
Figure 3 shows the effect of lack of centralisation on apparent Young’s modulus for a cylindrical
sector; this case approximates closely to the ASTM curved wall sample geometry (ref. Figure
A2.3 of API 5CT). For high D/t, the results converge to the simpler case of a rectangular cross-
section. All the production test results fall within −15.0, +23.9% of the NPL results (Table 2). The
right-hand plot of Figure 3 shows that these calculated values can be caused by sample
eccentricities of 2.7-3.9%, which are easily small enough to be credible, especially given curved
wall geometry.
Table 1 gives the stress-strain curve shape (SSCS) results, as Needleman’s n. It only gives data
for manufacturer XY01; those for manufacturer CT03 are inapplicable, because the NPL samples
were stress relieved.
There is almost no similarity between the mill and NPL results, with differences from 20 to
68190%. However, n for sample T20516-1 is very high (24900), and Figure 2b shows that any
response above n = 100 is effectively sharp-kneed. Taking this as the upper limit, the difference
for this sample reduces to 174%, and we can say that in general, the potential error in SSCS from
single-sided extensometry is in the order of hundreds of percent.
The NPL results show that three of the XY01 samples were rounded-kneed (n < 10); four were
sharp-kneed (n > 100); and three were between the two. If this small dataset is representative of
non-Q&T product as a whole (and it remains to be shown that it is), then it is wrong to say that
non-Q&T pipe will necessarily have either a rounded or sharp-kneed response; it appears that
either is possible, depending on production conditions.
4
Strictly, E111 requires a minimum of two standard strain gauges, or one averaging gauge; but the
overall effect is the same.
6.0 CONCLUSIONS
2) The errors in the latter are primarily due to lack of centralisation, and thus inaccurate strain
measurement for single-sided extensometry.
3) High-specification testing (to ASTM E111 or better) is required for accurate determination of
Young’s modulus and stress-strain curve shape.
7.0 RECOMMENDATIONS
1) Further testing is required in order to give data which can be used in design code calibration. The
present results are illustrative, but not statistically significant.
8.0 REFERENCES
1) “Formulas and templates for establishing casing, tubing, drill pipe and line pipe properties”, ISO
TR10400, ISO working draft, January 2007.
2) Huang, N.C. and Pattillo, P.D., “Collapse of oil well casing”, J. Pressure Vessel Technology,
ASME, vol. 104, February 1982.
3) Yeh, M.K. and Kyriakides, S., “On the collapse of inelastic thick-walled tubes under external
pressure”, J. Energy Resources Technology, ASME, vol. 108, March 1986.
4) “Collapse: development of non-Q&T ULS model (4)”, report prepared by Amerada Hess for
API/ISO TC67/SC5/WG2b, October 2002.
5) “Collapse: development of non-Q&T ULS model (5)”, report prepared by Amerada Hess for
API/ISO TC67/SC5/WG2b, January 2003.
6) “Collapse: effect of stress-strain curve shape (2)”, report prepared by Amerada Hess for API/ISO
TC67/SC5/WG2b, May 2001.
7) “Collapse: stress-strain curve shape for non-Q&T pipe”, report prepared by Amerada Hess for
API/ISO TC67/SC5/WG2b, June 2001.
8) “Collapse: development of non-Q&T ULS model”, report prepared by Amerada Hess for API/ISO
TC67/SC5/WG2b, January 2002.
9) “Collapse: development of non-Q&T ULS model (2)”, report prepared by Amerada Hess for
API/ISO TC67/SC5/WG2b, April 2002.
10) “Collapse: interim calibration for non-Q&T pipe”, report prepared by Amerada Hess for API/ISO
TC67/SC5/WG2b, May 2003.
11) “Specification for casing and tubing”, API Specification 5CT, 6th edition, October 1998.
12) “Standard test methods and definitions for mechanical testing of steel products”, ASTM standard
A370, 1997.
13) “Standard test methods for tension testing of metallic materials”, ASTM standard E8, ASTM,
1999.
14) “Standard test method for Young’s modulus, tangent modulus, and chord modulus”, ASTM
standard E111, 1997.
15) Roebuck, B. et al., “Data acquisition and analysis of tensile properties for metal matrix
composites”, J. Testing and Evaluation, ASTM, Vol. 22, no. 1, January 1994.
16) Excel spreadsheet containing manufacturer CT03 NPL stress-strain results, National Physical
Laboratory, December 2004.
17) Needleman, A. “Post-bifurcation behaviour and imperfection sensitivity of elastic-plastic circular
plates”, Int. J. Mechanical Sciences, vol. 17, 1975.
Young’s modulus
Sample API yield stress (ksi) Needleman’s n (-)
(ksi)
Test Prod. NPL Prod. Prod. NPL Diff. Prod. NPL Diff.
Calc. WG2B WG2B Mill WG2B WG2B (%) WG2B WG2B (%)
T20383 31,370 29,960 92.29 92.35 89.73 -2.8 10.23 12.25 19.7
T20383-1 24,170 29,950 89.26 87.76 89.20 -0.1 4.57 6.19 35.4
T20383-2 32,240 30,070 88.72 88.54 88.68 0.0 4.19 6.24 48.9
T20515 29,590 29,940 66.34 66.36 64.99 -2.0 24.81 226.4 813
T20515-1 33,360 29,820 64.79 64.75 66.31 2.3 34.68 222.9 543
T20515-2 30,980 29,950 65.62 65.60 65.74 0.2 226.8 152.3 -32.8
T20516 29,900 30,070 61.97 62.04 60.69 -2.1 17.51 4.04 -76.9
T20516-1 30,870 30,030 63.83 63.81 64.03 0.3 36.46 24900 68194
T20516-2 35,280 29,980 62.82 62.86 63.75 1.5 13.31 48.96 268
T20516-3 29,030 29,780 64.29 64.26 62.65 -2.6 314.1 27.21 -91.3
Mean 30,679 29,955
COV 0.0961 0.0032
Notes
Stress (ksi)
1 2
50 fpbyfy = 0.717 50 fpbyfy = 0.641
1 2
E = 28100 ksi E = 34780 ksi
1 2
n = 8.05 n = 6.51
1 2
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
3 4
50 fpbyfy = 0.658 50 fpbyfy = 0.707
3 4
E = 31430 ksi E = 28340 ksi
3 4
n = 7.18 n = 8.23
3 4
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
5 6
50 fpbyfy = 0.695 50 fpbyfy = 0.628
5 6
E = 32970 ksi E = 31030 ksi
5 6
n = 8.29 n = 5.87
5 6
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
7 8
50 fpbyfy = 0.736 50 fpbyfy = 0.682
7 8
E = 33380 ksi E = 30410 ksi
7 8
n = 9.78 n = 7.88
7 8
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
9 10
50 fpbyfy = 0.521 50 fpbyfy = 0.688
9 10
E = 24940 ksi E = 28840 ksi
9 10
n = 3.14 n = 7.36
9 10
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Fitted Fitted
Actual Actual
Figure 1a: Stress-strain curve shape results, manufacturer CT03, J55 (production tensile test data)
Pipe 2921 Pipe 2922
100 100
fy = 67.47 ksi
1
fp = 67.3 ksi
Stress (ksi)
Stress (ksi)
1
50 fpbyfy = 0.998 50
1
E = 29830 ksi E = 29800 ksi
1 2
n = 3269
1
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
3 4
50 fpbyfy = 0.998 50 fpbyfy = 1
3 4
E = 30210 ksi E = 29810 ksi
3 4
n = 3313 n = 38120
3 4
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
5 6
50 fpbyfy = 0.998 50 fpbyfy = 0.999
5 6
E = 30010 ksi E = 29980 ksi
5 6
n = 4290 n = 17180
5 6
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
7 8
50 fpbyfy = 0.995 50 fpbyfy = 0.999
7 8
E = 29760 ksi E = 29990 ksi
7 8
n = 1428 n = 6703
7 8
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
9 10
50 fpbyfy = 1 50 fpbyfy =1
9 10
E = 29700 ksi E = 29620 ksi
9 10
n = 123800 n = 21700
9 10
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Fitted Fitted
Actual Actual
Figure 1b: Stress-strain curve shape results, manufacturer CT03, J55 (NPL tensile test data)
Sample T20383 (K55 HC) Sample T20383-1 (K55 HC)
100 100
fy = 92.35 ksi fy = 87.99 ksi
1 2
fp = 71.80 ksi fp = 59.59 ksi
Stress (ksi)
Stress (ksi)
1 2
50 fpbyfy = 0.777 50 fpbyfy = 0.677
1 2
E = 31370 ksi E = 24670 ksi
1 2
n = 10.23 n = 4.53
1 2
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
3 4
50 fpbyfy = 0.577 50 fpbyfy = 0.862
3 4
E = 32240 ksi E = 29590 ksi
3 4
n = 4.19 n = 24.81
3 4
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
5 6
50 fpbyfy = 0.886 50 fpbyfy = 0.975
5 6
E = 33360 ksi E = 30980 ksi
5 6
n = 34.68 n = 226.8
5 6
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
7 8
50 fpbyfy = 0.816 50 fpbyfy = 0.892
7 8
E = 29900 ksi E = 30870 ksi
7 8
n = 17.51 n = 36.46
7 8
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
9 10
50 fpbyfy = 0.763 50 fpbyfy = 0.981
9 10
E = 35280 ksi E = 29030 ksi
9 10
n = 13.31 n = 314.1
9 10
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Fitted Fitted
Actual Actual
Figure 2a: Stress-strain curve shape results, manufacturer XY01 (production tensile test data)
Sample T20383 (K55 HC) Sample T20383-1 (K55 HC)
100 100
fy = 89.73 ksi fy = 89.20 ksi
1 2
fp = 72.28 ksi fp = 61.64 ksi
Stress (ksi)
Stress (ksi)
1 2
50 fpbyfy = 0.805 50 fpbyfy = 0.691
1 2
E = 29960 ksi E = 29950 ksi
1 2
n = 12.25 n = 6.19
1 2
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
3 4
50 fpbyfy = 0.690 50 fpbyfy = 0.975
3 4
E = 30070 ksi E = 29940 ksi
3 4
n = 6.24 n = 226.4
3 4
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
5 6
50 fpbyfy = 0.975 50 fpbyfy = 0.966
5 6
E = 29820 ksi E = 29950 ksi
5 6
n = 222.9 n = 152.3
5 6
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
7 8
50 fpbyfy = 0.495 50 fpbyfy = 1.000
7 8
E = 30070 ksi E = 30030 ksi
7 8
n = 4.04 n = 24900
7 8
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Stress (ksi)
9 10
50 fpbyfy = 0.915 50 fpbyfy = 0.867
9 10
E = 29980 ksi E = 29780 ksi
9 10
n = 48.96 n = 27.21
9 10
0 0
0 0.5 1 0 0.5 1
Strain (%) Strain (%)
Fitted Fitted
Actual Actual
Figure 2b: Stress-strain curve shape results, manufacturer XY01 (NPL tensile test data)
Bending +ve, strain at convex face Do = 4.5", Do/t = 20
3 3
Apparent E/actual E (-)
1 1
0 0
10 5 0 5 10 10 5 0 5 10
Eccentricity/sample thickness (%) Eccentricity/sample thickness (%)
Do = 4.5", Do/t = 10 Bending +ve, convex face
Do = 4.5", Do/t = 40 Bending +ve, concave face
Do = 20", Do/t = 10 Bending -ve, convex face
Do = 20", Do/t = 40 Bending -ve, concave face
File note: Test programme for determination of stress-strain curve shape for non-Q&T product
1.0 Background
API/ISO SC5/TC67 WG2B have developed interim collapse ratings for non-Q&T pipe [1], based
on adjustment of the Q&T ratings for the additional loss in strength due to stress-strain curve
shape (SSCS) effects. These interim ratings have been published in the recent DIS issue of ISO
10400 [2]. However, the approach used suffers from several limitations, as follows:
• The collapse tests were all for J/K55 pipe. The behaviour for other non-Q&T grades may
be different.
• The tensile test results for each collapse test sample were not retained, and therefore it
was not possible to determine the SSCS in each case1. The ratings are therefore for the
average SSCS, and it is possible that the strengths for highly rounded-kneed stress-
strain curves are considerably lower than those given.
• The present non-Q&T model does not include for the dispersion of collapse strength due
to SSCS. Therefore, the calculated ratings are likely to be unconservative, perhaps
seriously so if SSCS varies widely.
• Present test data suggests that the shape of the decrement (“dip”) curve used for non-
Q&T collapse strength prediction is incorrect.
Because of this, the present ratings are not considered adequate for industry use, and it is
proposed that design strengths be calculated using a structural reliability technique. This method
has already been used for calculation of the ratings for Q&T pipe.
Application of the structural reliability method to non-Q&T pipe requires both development of a
suitable analysis model, and determination of material property data for SSCS. This file note
briefly describes the work proposed in each area, and gives a detailed discussion (and
justification) of the SSCS test programme.
1) There is no reliable data on SSCS for non-Q&T product. This is because production tensile tests
do not measure axial strain with sufficient accuracy [3]. The present workscope includes a
comparison exercise to demonstrate this conclusively.
2) There is no validated analysis model for non-Q&T product. While continuum mechanics models
have been developed by Huang and Pattillo [4] and Yeh and Kyriakides [5], neither was subject
to comprehensive verification against collapse test data; and the few data points that were used
are open to question, given present understanding of the limitations of production-quality tensile
tests. In any event, neither model is presently available to the committee.
A new analysis model, using finite element (FE) methods, is currently in preparation, and should
be completed and validated by the June 2005 meeting.
3) The present residual stress data is open to question. This is because the test samples were
generally taken from the pipe end, which does not see the full effect of rotary straightening.
3.0 Questions
1
In any event, it proves that production-quality tensile testing does not measure axial strain with
sufficient accuracy to allow calculation of SSCS, see Section 2.0.
1 of 5
Proposal: test programme using high-accuracy strain measurement (this programme).
Proposal: specify measurement of residual stress at pipe centre. This will require the use of an
NDE technique in order to be acceptable to the industry. Until recently, no such method was
available, but a recently developed approach using changes in magnetism shows considerable
promise [6]. It will, however, require verification before any adoption by the industry.
4.0 Objective
The programme objective is to determine the statistical variation of stress-strain curve shape for
non-Q&T product. This will enable the calculation of non-Q&T design collapse strengths once the
analysis model is complete.
In order to understand the test programme design, it is first necessary to understand the data
input for the FEA code, and how this data is used. The element bending stiffness is obtained, as
a function of axial strain and curvature, by integrating the inelastic stress-strain response over
the wall thickness. The wall thickness is divided into 100 elements, with 20-point Gaussian
integration within each element. The stress-strain response is thus evaluated at some 2000
positions through the wall thickness. In general, the strain (and therefore the stress) will be
different at each position.
The residual stress is modelled explicitly, as an initial stress at each position. The total stress at
each position is thus the initial stress plus the base stress (that is, the material intrinsic stress-
strain response) for the current strain. The initial stress at any given position thus defines an
offset curve to the base stress-strain response. Therefore, it is the base curve (that is, the
intrinsic stress-strain response, for no residual stress) which must be determined by the test
programme. It follows that the samples must be stress relieved before testing, to remove any
residual stress present.
In principle, the FEA code could be used to predict the stress-strain response for the sample as a
whole, for a given residual stress, and the prediction compared to the actual response from test
data. In practice, however, there are two difficulties with this approach. First, calculation of the
global stress-strain response requires consideration of the sample lengthwise behaviour, which
is not yet well understood. Second, comparison against experimental results involves testing of
full wall-thickness (curved surface) samples, which is not possible because high-accuracy strain
measurement requires testing of rectangular cross-section samples cut from within the wall
thickness.
The data set-up for the FEA model assumes that the intrinsic stress-strain response is constant
over the wall thickness. In practice, this is not the case, as yield stress varies slightly throughout
the wall thickness (see, e.g., [7]), because of minor variations in production parameters
(tempering temperature, degree of yielding during straightening (Bauschinger effect), and so on.)
Similarly, it will be assumed that residual stress varies linearly over the wall thickness, which
again is an idealisation.
These idealisations are necessary because it is not practically possible to model the actual
variation of each quantity through the wall thickness, even if it were known. Even were such
measurement possible, it would in turn raise further questions about the statistical
characterisation of each variation, which again will be complex.
It is therefore suggested that the best way forward is to tolerate the idealisations in the FEA
model, perhaps running some sensitivities to check the degree of likely variation in collapse
strength, and to investigate the question of accuracy of fitting via comparison against collapse
test results, as already done for the Q&T model. This will however require a catalogue of suitable
collapse tests; this does not exist at present, and it is likely that gathering suitable test results will
take some time (perhaps years).
In summary:
2 of 5
• The FEA approach requires use of the intrinsic stress-strain response (i.e., without
residual stress), together with idealisation of the variation of yield and residual stress
through the wall thickness.
• The need for high-accuracy strain measurement, and the practical application of the
analysis method, preclude any more complex approach.
Twenty (20) samples will be tested. This is the maximum allowed by the available funds, but is
rather below the number required for statistical significance. If the SSCS behaviour proves to be
homogeneous, then the results from the present test series may just suffice; but it is more likely
that a further funding application will be made, based on the learnings from the present
programme.
Accuracy of strain measurement will comply with or exceed the provisions of ASTM E111 [8].
This level of accuracy is beyond the reach of normal equipment, and requires the services of a
specialist materials testing company. The UK National Physical Laboratory (NPL) has been
chosen for this purpose.
Testing will be conducted in two batches of ten (10) samples each, to enable learnings from the
first test series to be fed back to WG2B, and the experiment design for the second test series
revised if appropriate. The interval between the two test series will depend on the timing of the
next WG2B meeting, but is likely to be between one and three months.
Rough-cut sample blanks will be provided, cut from the pipe wall curved surface, for preparation
of tensile test coupons by NPL. Sample blanks will be approximately 120 mm (pipe lengthwise
direction) x 40 mm (circumferential direction), and will be cut by bandsaw or similar methods (i.e.,
not flame cut). Samples will be stress relieved by NPL before testing.
Each blank will be marked up with a unique sample ID by the manufacturer (see 7.0 below).
Sample identity shall be retained throughout coupon preparation and tensile testing, and the
manufacturer sample ID given with each set of tensile test results. This is required to allow
comparison of the NPL and manufacturer tensile test results for each pipe.
The deliverable from each test series will be the stress-strain curves, as digital X-Y data in Excel
format. A brief report may also be provided if desired.
1) A pair of samples shall be cut from adjacent positions on the pipe. The first shall be tensile tested
by the manufacturer, to the standard API 5CT/ASTM A370 test specification [9, 10] (i.e., the
manufacturer’s default process for production monitoring). The second shall be tested by NPL to
a high-accuracy strain measurement specification [8]. To allow comparison of manufacturer and
NPL test results, each pair of samples shall be marked up with a unique sample ID.
2) Twenty (20) pairs of samples will be tested, all for grade J/K55 cold rotary straightened product.
They may be made up of either 20 pairs from one mill, or 10 pairs from each of two mills.
3) Manufacturer tensile test results shall be provided as digital X-Y data in Excel format. The
following information shall be provided with each set of tensile test results:
3 of 5
The manufacturer shall ensure that the tensile testing equipment is within the documented
quality assurance program certification period.
To protect confidentiality, manufacturers may withhold their identity if desired. Regardless of this,
all results reporting to API etc. will be non-attributable.
6) The sample blanks provided to NPL shall be approx. 120 mm (pipe lengthwise direction) x 40
mm (circumferential direction), for preparation of tensile test coupons by NPL. In order to allow
cutting of a rectangular cross-section test coupon from within the curved surface wall thickness,
the OD and weight of the parent pipe shall be as follows:
Sample blanks shall be cut using band-saw, abrasive rotating blade, carbide or high-speed
tooling methods.
7) Pipes may be defect rejects (pits, crack-like imperfections, etc.), but should be otherwise
representative of the production process. If defect reject pipe is used, the sample blanks should
not contain the defect(s).
8) Phase I samples may be taken from any position along the pipe, (i.e., pipe end samples are
acceptable). Phase II samples shall be taken from within the length subject to the full effect of
rotary straightening (i.e., inboard of the second roll from either end of the joint).
8.0 References
1) “Collapse: interim calibration for non-Q&T pipe”, report prepared by Amerada Hess for API/ISO
TC67/SC5/WG2b, May 2003.
2) “Formulas and templates for establishing casing, tubing, drill pipe and line pipe properties”, ISO
10400, DIS issue, October 2003.
3) “Collapse: development of non-Q&T ULS model (5)”, report prepared by Amerada Hess for
API/ISO TC67/SC5/WG2b, January 2003.
4) Huang, N.C. and Pattillo, P.D., “Collapse of oil well casing”, J. Pressure Vessel Technology,
ASME, vol. 104, February 1982.
4 of 5
5) Yeh, M.K. and Kyriakides, S., “On the collapse of inelastic thick-walled tubes under external
pressure”, J. Energy Resources Technology, ASME, vol. 108, March 1986.
6) Thayer, P., “MAPS – non-destructive measurement of stress in oilfield tubulars”, presentation to
SPE Aberdeen chapter, January 2004.
7) Thoft-Christensen, P. and Baker, M.J., Structural reliability theory and its applications, Springer-
Verlag, 1982.
8) “Standard test method for Young’s modulus, tangent modulus, and chord modulus”, ASTM
standard E111, 1997.
9) “Specification for casing and tubing”, API Specification 5CT, 6th edition, October 1998.
10) “Standard test methods and definitions for mechanical testing of steel products”, ASTM standard
A370, 1997.
5 of 5
API/ISO TC67/SC5/WG2b: Modernisation of API Bulletin 5C3
API PRAC Report: Determination of Stress-Strain Curve Shape for J/K55 Product
(2⋅vec6⋅ε i + vec5)⋅psi ( )
2
var ← Ej − µ
k k
minj j←j+1
minj
get_maxj( minj , tanE) := j ← minj + 1
while tanE > llim_E⋅ 30000 ⋅ ksi get_minj( tanE) := vec ← ( get_minj1( tanE) get_minj2( tanE) )
j
max( vec)
maxj ← j
Tangent modulus
j←j+1
i := n2 + 1 .. n1 − n2 tanE := get_tanmod( i)
maxj i
j := n2 + n3 + 1 .. n1 − n2 − n3
Choose locally flattest response
var_tanE := get_var( tanE , n3 , j)
get_bestj( var , minj , maxj) := minvar ← min( submatrix( var , minj , maxj , 1 , 1 ) ) j
j ← minj minj := get_minj( tanE)
−1
Adjust for δε := ε bestj − σ bestj⋅ tanmod ε1 := ε − δε
grip slippage
i2 := 1 .. rows( ε )
−1
hε := ε1 minj⋅ ( minj − 1 ) k := 1 .. minj − 1 ε1 k := ( k − 1 ) ⋅ hε σ k := ε1 k⋅ tanmod
bestj := get_bestj( var_secE , minj + n4 + 1 , maxj − n4) E := secE σE1 := ε1 ⋅ E E = 28101 ksi
bestj
4 .10
4
100
3 .10
4
−1 −1
tanEi⋅ ksi σ ⋅ ksi
2 .10
4
50
−1 −1
secEi⋅ ksi σE1⋅ ksi
1 .10
4
0 0
0 20 40 60 0 0.5 1
−1 −1
σ i⋅ ksi ε1⋅ %
yd ← 0 σd := σd ⋅ psi
1
for i ∈ 2 .. 1000 100
xd ← xd + hx
i i− 1
−1
σ ⋅ ksi
( )
−1
yd ← interp x , y , xd ⋅ psi
i i −1
σE1⋅ ksi 50
return ( xd yd ) if xd + hx > max( x )
i −1
σd⋅ ksi
return ( xd yd ) if xd + hx > maxx
i
σbelow( ε , E) := ε ⋅ E
−1
n n := 5 σp := 50⋅ ksi
σabove( ε , σp , E , n ) := σp ⋅ ⎛⎜ n ⋅ ε ⋅ + 1 − n⎞
E
⎝ σp ⎠ getn( σy , εy , σp , E) := root( σabove( εy , σp , E , n ) − σy , n )
n3 := 101 j := 1 .. n3 εp j :=
1.5%
n3 − 1
⋅ ( j − 1) (
σpr1 j := σpr εp j , fp , E , n )
:= rows( ε1 )
−1 −1
results results := εp j results := σpr1 j⋅ ksi results := ε1 i2 results := σ i2⋅ ksi
6, 1 j, 2 j, 3 i2 , 4 i2 , 5
α ( R) := asin⎛⎜
0.5⋅ in ⎞ R
Dimensions t( R , Rbyt) := e( R , Rbyt , ebyt) := ebyt⋅ t( R , Rbyt) ORIGIN ≡ 1
⎝ R ⎠ Rbyt
2
⋅ ⎛⎜ 2 − ⎞
α ( R) ⋅ R 1 This approximates a curved sample to API 5CT Section 9.8.2
Section A( R , Rbyt) :=
properties Rbyt ⎝ Rbyt ⎠ Sample throat width (dimension A of Figure A2.3) = 1"
⎡4 ⎛ 2 ⋅ sin( α ( R) ) ⎞
2 ⎤
I( R , Rbyt) :=
R ⎢ ⎛
⋅ 1−
3
+
1
−
1 ⎞ ( ) ( )
⋅ ⎜ α ( R) + sin α ( R) ⋅ cos α ( R) − ... ⎥
Rbyt ⎢ ⎜ 2 ⋅ Rbyt 2 3 ⎝ α ( R) ⎠ ⎥
⎢⎝ ⎠
Rbyt 4 ⋅ Rbyt
⎥
sin( α ( R) ) ⎛
2
⎢+ 1 1 1 ⎞ ⎥
⋅ ⋅ 1− +
⎢ 2
⎣ 3⋅ Rbyt ⋅ 2 − Rbyt (−1 α ( R) )
⎜
⎝
Rbyt 2
6 ⋅ Rbyt ⎠
⎥
⎦
⎡ 2 ⋅ sin( α ( R) ) ⎛ 1 1 ⎞ ⎤ (Convex face)
y1a( R , Rbyt) := R⋅ ⎢ 1 − ⋅ 1− + ⎥
3⋅ α ( R) ⎜ Rbyt − 1
⎣ ⎝ 2 − Rbyt ⎠ ⎦
y1b ( R , Rbyt) := t( R , Rbyt) − y1a( R , Rbyt) (Concave face, note definition of y1b is
different from that in Roark)
Axial stress per unit axial load, linear response
−1 −1
σax( R , Rbyt) := A( R , Rbyt) n := 100 i := 1 .. n ebyt := −0.1 + 0.2⋅ ( i − 1 ) ⋅ ( n − 1 )
i
Extreme fibre stresses per unit axial load, linear response
−1
σb +ve on side a σefap( R , Rbyt , ebyt) := σax( R , Rbyt) + 1 ⋅ e( R , Rbyt , ebyt) ⋅ y1a( R , Rbyt) ⋅ I( R , Rbyt)
−1
σefbp( R , Rbyt , ebyt) := σax( R , Rbyt) − 1 ⋅ e( R , Rbyt , ebyt) ⋅ y1b ( R , Rbyt) ⋅ I( R , Rbyt)
−1
σb -ve on side a σefam( R , Rbyt , ebyt) := σax( R , Rbyt) − 1 ⋅ e( R , Rbyt , ebyt) ⋅ y1a( R , Rbyt) ⋅ I( R , Rbyt)
−1
σefbm( R , Rbyt , ebyt) := σax( R , Rbyt) + 1 ⋅ e( R , Rbyt , ebyt) ⋅ y1b ( R , Rbyt) ⋅ I( R , Rbyt)
2 2
1 1
0 0
10 5 0 5 10 10 5 0 5 10
Eccentricity/sample thickness (%) Eccentricity/sample thickness (%)
Do = 4.5", Do/t = 10 Bending +ve, convex face
Do = 4.5", Do/t = 40 Bending +ve, concave face
Do = 20", Do/t = 10 Bending -ve, convex face
Do = 20", Do/t = 40 Bending -ve, concave face