13th International Conference on Fracture

June 16-21, 2013, Beijing, China

API 579 G-factors for K Calculations and Improvements For Assessment of

Crack-like Flaws in Pipelines

Shaikh Rahman, Ming Gao*, Ravi Krishnamurthy

Blade Energy Partners, Houston, TX, USA
Corresponding author: [email protected]

Abstract
API 579-1/ASME FFS-1 2007 gives a complete guideline of Fitness-For-Service (FFS) assessment of crack-
like flaws in structural components based on the Failure Assessment Diagram (FAD) method. The
assessment is carried out in FAD using a non-linear relationship between the load ratio (Lr) and the
toughness ratio (Kr) where the toughness ratio (Kr) is defined as the ratio of the Mode I stress intensity factor
(KI) to the toughness of the material (Kmat). One of the methods for Mode I stress intensity factor (KI)
calculation in API 579-1/ASME FFS-1 2007 is FEA based, which correlates five influence coefficients,
namely Go to G4, for KI calculation. These influence coefficients are dimensionless and are tabulated for a
range of two ratios; the wall thickness to the internal radius (t/Ri) and the flaw depth to the wall thickness
(a/t) for infinite cracks and three ratios (crack depth to half crack length (a/c) besides a/t and t/Ri) for finite
cracks. API 579-1/ASME FFS-1 2007 allows interpolation of the influence coefficients, Gs for intermediate
values. Even though it is not explicitly mentioned about the method of interpolation, a linear interpolation is
generally considered. Another possible method is explored here with the polynomial function fitted by the
FEA data provided in the table given in API 579. A detail discussion on obtaining the influence coefficients
for an intermediate value is outlined here. Case studies are provided to illustrate the advantages and
disadvantages of the linear interpolation and the fitted polynomial function.

Keywords: Stress Intensity Factor, Influence Coefficients Gi, G Interpolation, FAD, Level 2 Limiting
Pressure

1 Introduction
The Mode I Stress Intensity Factor (KI) in API 579 is a fourth order polynomial function of /
or / with applied internal pressure and pipe dimension. The solution is obtained from FEA
work completed by T. L. Anderson Group [1]. According to API 579, the Mode I Stress Intensity
Factors for the infinite length internal and external surface cracks in axial (longitudinal) direction
are given by the following equations [2] respectively.

Internal axial infinite surface crack:

2 2 3 4 5 √

External axial infinite surface crack:

2 2 3 4 5 √

-1-
Where, p is the internal pressure, Ri and Ro are the internal and external radii respectively, a is the
flaw depth and Go to G4 are the influence coefficients. This equation is valid for 0.0 / 0.8
and 0.001 / 1.0. The influence coefficients are given in Table C.10 and Table C.11 in API
579 document for infinite axial and circumferential cracks respectively. The table gives values of
influence coefficients for various t/Ri and a/t. t/Ri is converted to Ri/t for the convenience in
calculation performed in this work. Ri/t can also be readily converted to OD/t [ 2 / 1 ] that
is more common in tubular performance calculation in Oil and Gas Industry. For finite surface
crack √ in the above two equations are replaced by / . The influence coefficients for the
finite crack are function of a/c in addition to a/t and t/Ri ratios. Therefore, for infinite crack, two
interpolations and for finite crack, three interpolations are required to obtain the desired influence
coefficients. It is mentioned in API 579 that “interpolation of the influence coefficients, Gi, may be
used for intermediate values of t/Ri and a/t” but does not state explicitely how the interpolation
should be performed, linearly or otherwise. The work described in this paper takes an attempt to
explain if the linear interpolation is adequate or accurate enough for this purpose. The cubic
polynomial interpolation is used for the comparative study.

2 Interpolation methods
2.1 Infinite Surface Crack
Table C.10 and Table C.11 in API 579 give the influence coefficients of infinite surface crack for
axial and circumferential direction respectively. Figure 1 shows part of the Table C.10 and Table
C.11 from API 579. The influence coefficients shown in these tables are function of t/Ri and a/t.
Therefore, two interpolations are required to obtain the desired influence coefficients for a given
pipe and crack dimensions.

Figure 1 – Partial Table C.10 and Table C.11 from API 579

Interpolation is performed in two steps using the following methodology:

 For a given pipe and crack geometry, a/t and t/Ri are calculated.
 The a/t and t/Ri are then used to identify the neighboring a/t and t/Ri in the table. There are
two sets of influence coefficients for each t/Ri obtained from two neighboring a/t.
 First interpolation is performed for each t/Ri between the corresponding set of a/t for the
desired a/t.

-2-
  The second and final interpolation is performed between those interpolated influence
coefficients for the desired t/Ri.

The order of the interpolations (starting with a/t or t/Ri) is irrelevant. Linear interpolation requires
two sets of data points surrounding the desired a/t or t/Ri. But for the cubic polynomial interpolation
requires four sets of data points. In the subsequent sections a detail description is given with
examples on how to perform the both interpolations and is shown the difference with respect to
linear interpolation. Out of the five influence coefficients, the first one, Go has the most influence in
the stress intensity function. Therefore, the comparative analysis is performed on Go alone.

2.1.1 Nature of nonlinearity

The influence coefficients are function of a/t and inverse of Ri/t. The nature of nonlinearity of the
influence coefficients with a/t and Ri/t are discussed here starting with a/t.

Figure 2 shows the nature of non-linearity of all five influence coefficients for Ri/t = 10. The first
influence coefficient shows the highest nonlinearity and the nonlinearity diminishes for the
subsequent influence coefficients. It also shows a very good fit with cubic polynomial. For other
Ri/t the cubic polynomial function will change. Figure 3a shows the first influence coefficient (Go)
for Ri/t = 10 and Ri/t = 5. It can be visually inferred from Figure 3a that linearity can be assumed
between any two consecutive points without losing a considerable amount of accuracy. The effect
of non-linearity on Go can be observed from Figure 3b. Non-linearity reduces with lesser Ri/t as
well as lesser a/t.

There are seven points identified in the Figure 3a. They are -

 Point A: a/t=0.6 and Ri/t=5

 Point B: a/t=0.6 and Ri/t=10
 Point C: a/t=0.8 and Ri/t=5
 Point D: a/t=0.8 and Ri/t=10
 Point E: a/t=0.7 and Ri/t=5. This is one of the two interpolated points obtained after the first
interpolation.
 Point F: a/t=0.7 and Ri/t=10. This is one of the two interpolated points obtained after the
first interpolation.
 Point X: a/t=0.7 and Ri/t=7. This is the final value after the second interpolation between
points E and F.

Point A through Point D are obtained from Table C.10. The example discussed later uses these
points and shows the differences in calculating Go using cubic polynomial interpolation and linear
interpolation. Next discussion is on the nonlinear nature of Go with respect to Ri/t.

Figure 4a shows the nature of the nonlinearity of Go with respect to Ri/t. It is observed from the
earlier figures that the non-linearity reduces with decreasing a/t as well as with decreasing Ri/t.
Even more interestingly, Figure 4a shows that the nonlinearity of Go vanishes beyond a value of
Ri/t. This value varies with variation of a/t. Assuming the maximum Ri/t = 40 after which the
influence of Ri/t vanishes, Figure 4b shows a set of cubic polynomials of Ri/t for various a/t. These
are not as good fit as we have seen earlier with cubic polynomial of a/t even though the maximum

-3-
of Ri/t is reduced down to 40. Improvement can be made in the curve fitting is to perform the curve
fitting around the desired value. In explaining this, two piecewise curve fittings are performed and
shown in Figure 5. Curve fittings with Ri/t from 1 to 10 (left in the figure) and from 10 to 40 (right
in the figure) are considered in this figure.

Axial Internal Infinite Crack for Ri/t = 10

6
G0
y = 6.6905x3 + 0.6426x2 + 0.7337x + 1.1173
G1 R² = 1
5 G2 y = 2.2156x3 + 0.4336x2 + 0.2584x + 0.6809
G3 R² = 1
G4
4 y = 1.1111x3 + 0.307x2 + 0.1364x + 0.5239
Poly. (G0)
R² = 0.9999
G values
Poly. (G2)
3 Poly. (G2)
y = 0.6789x3 + 0.2208x2 + 0.0892x + 0.44
Poly. (G4) R² = 0.9999
Poly. (G4)
2
y = 0.4344x3 + 0.2184x2 + 0.06x + 0.3787
R² = 0.9999

0
0 0.2 0.4 0.6 0.8 1

a/t

Figure 2 – Nonlinear behavior of the influence coefficients of the axial internal infinite surface crack
for t/Ri = 10

Figure 3 – (3a) First influence coefficient (Go) for t/Ri = 10 and 5 and (3b) Effect of non-linearity in Go
with increasing Ri/t

Figure 6 shows the similar arrangement of points A through D described earlier along on the cubic
curve fit between Ri/t of 1 to 10. It also shows the linear approximation lines. It can be clearly
inferred from this figure that linearity can be assumed between two adjacent points as shown in the
previous case.

-4-
 Figure 4 – (4a) The first influence coefficient (Go) with respect to Ri/t and (4b) Nonlinear behavior of
the first influence coefficient with respect to Ri/t

Figure 5 – Local cubic polynomial of the first influence coefficient with respect to Ri/t. (a) Ri/t ≤ 10 (left
figure) (b) 10 ≤ Ri/t ≤ 40 (right figure)

Go (Axial Internal Infinite Crack)

6
a/t=0.6 D
X
a/t=0.8
a/t = 0.7
a/t=0.7 Ri/t =7
5
Poly. (a/t=0.6)
F
Poly. (a/t=0.8) C
Poly. (a/t=0.7)
4

E
Go

3
B

2 A

y = 0.0014x3 ‐ 0.0419x2 + 0.7107x + 1.2641

1 R² = 1
y = ‐0.0063x2 + 0.2834x + 2.0354
y = 0.004x3 ‐ 0.0866x2 + 0.6901x + 0.9435 R² = 1
R² = 1
0
0 5 10

Ri/t

Figure 6 – First influence coefficient (Go) for a/t = 0.6 and 0.8

-5-
Therefore, linearity between two adjacent a/t points and two Ri/t points can be assumed without
losing much accuracy. Following example takes an attempt of quantifying the accuracy. The first
influence coefficient (Go) is selected because it has the most influence in stress intensity function
(Equation is given at the beginning).

2.1.2 Example
Let us assume that the influence coefficients for a/t=0.7 and Ri/t=7 are needed for calculating stress
intensity factor. Table 1 shows a portion of Table C.10 from API 579 for Ri/t of 10 and 5 and for a/t
from 0 to 0.8. The values we are looking for are between Ri/t of 10 and 5 and a/t of 0.6 and 0.8.
These values are marked in the table. As it is mentioned earlier that two interpolations are required.
The first interpolation is performed for a/t = 0.7 that is in between a/t of 0.6 and 0.8. This yields two
sets of influence coefficients; one for Ri/t = 10 and the other for Ri/t = 5. The next and final
interpolation is performed for Ri/t = 7 that is in between Ri/t of 10 and 5. It is also important to note
that the interpolations are performed between values of a/t and Ri/t that are adjacent to the desired
values.

Table 1 – Influence coefficients for Axial Internal Infinite Crack

2.1.3 Interpolation between a/t

Following are couple of important notes that can be inferred from Figure 2:

 The first influence coefficient (Go) is more nonlinear than other influence coefficients.
 For lesser a/t, the function is more linear than higher a/t.
 A linear behavior may be assumed between two adjacent a/t.

Figure 3 shows the first influence coefficients for Ri/t = 10 and Ri/t = 5. Two vertical straight lines
are drawn between a/t = 0.6 and a/t = 0.8. Four points are identified on these two lines. The values
of the points are given in Table 2 along with the values computed using cubic functions and the
associated % difference. It is important to note that even at the known values of a/t and Ri/t the
cubic functions yield difference even though the % difference are minor.

Table 2 – Values of four points used in the interpolations

-6-
Next, we calculate Go for a/t = 0.7 by using cubic polynomial function and by using linear
interpolation method. Linear interpolation method uses the tabular values of Go for a/t = 0.6 and a/t
= 0.8. The points E and F in Figure 3 are the interpolated points for a/t = 0.7 for Ri/t = 5 and Ri/t =
10 respectively. Table 3 shows the values of points E and F by using cubic and linear interpolations.
The % difference shown in the table indicate that the linear interpolation yields 1.27% and 2.36%
higher Go in reference to the cubic interpolated values. It is important to remember that the values
using cubic functions associate with some difference.

Table 3 –Interpolation for a/t=0.7 between 0.6 and 0.8

2.1.4 Interpolation between Ri/t

Figure 6 shows influence coefficient points A through F on Ri/t axis. A cubic curve fit is obtained
for a/t=0.7. Since cubic polynomial interpolation requires four points to perform the calculation we
calculated Go for a/t=0.7 and Ri/t=20. Using the cubic polynomial function for a/t=0.7, Go is
calculated for Ri/t=7. Finally, linear interpolation is performed between Ri/t = 5 and 10 for Ri/t = 7
on a/t = 0.7. The final interpolated values for point X using cubic and linear interpolation are shown
in Table 4. The overall % difference is 1.22% for linear interpolated value in reference to the cubic
interpolated value.

Table 4 – Interpolation for Ri/t=7 between 5 and 10 at a/t=0.7

2.1.5 Findings of the example

 Cubic interpolation introduces complexity in the calculation without improving much
accuracy in the results.
 Besides computational complexity following are some comments and observations on cubic
polynomial and curve fitting:
o Cubic polynomial interpolation requires four consecutive points if local curve fitting
is intended otherwise global curve fitting may be employed.
o Global curve fitting introduces more difference in the interpolation.
o Cubic polynomial function needs to be obtained for each set of Ri/t and a/t that will
introduce complexity in the calculation and increase computation time.
 The % difference in reference to the cubic interpolation is reasonable if linear interpolation
is performed.
 Similar arguments for linear interpolation on other influence coefficients can be made.

-7-
2.2 Finite Surface Crack
Figure 7 shows partial tables for influence coefficients of finite axial inside and outside surface
cracks. Unlike the tables for infinite crack, these tables contain one more level of variable that is the
ratio of crack depth to the half crack length (a/c). Hence, to obtain the desired influence
coefficients, three interpolations are required.

Figure 7 – Partial Table C.12 and Table C.13 from API 579

Interpolation is performed using the following methodology:

 For a given pipe and crack geometry, t/Ri, a/c and a/t are calculated.
 The a/t a/c and t/Ri are then used to identify the neighboring a/t, a/c and t/Ri in the table.
There are two sets of influence coefficients with the neighboring a/t for each t/Ri and a/c.
 First interpolation is performed for each t/Ri and a/c between the corresponding set of a/t for
the desired a/t.
 The second interpolation is performed between those interpolated influence coefficients for
the desired a/c for each t/Ri.
 Finally the third and final interpolation is performed for the desired t/Ri.

2.3 Effect of interpolation on limiting pressure prediction

In the earlier section, we discussed and showed in the example the differences in the influence
coefficients when using the cubic polynomial interpolation instead of the linear interpolation. With
the finite crack we will calculate Level 2 limiting pressure using both interpolation methods. Level
2 limiting pressure for a known crack is defined as the pressure at which the toughness ratio (Kr)
equals to the toughness ratio calculated from the Level 2 FAD equation for the calculated load ratio
(Lr). This is the maximum operation pressure beyond which the crack will be unacceptable for
fitness for service according to Level 2 FAD. FAD is a relation between Kr and Lr, which is
described as follows:
.
1 0.14 0.3 0.7

-8-
Where, 1.25 for Carbon Steels, / and / , KI is the mode I stress
intensity factor, is the materials fracture toughness, is the reference stress, based on yield
load and limit load solutions for the configuration of interest, and is the yield strength of the
material. Figure 8 shows the FAD using the above equation. If the assessment point lies inside the
FAD, the structure is considered safe and the crack is acceptable for fit. The unacceptable crack is
predicted when the assessment point falls outside the FAD.

Figure 8 – FAD for Level 2 Assessment

Both KI and calculated for a given crack are function of operating pressure. The limiting
pressure is that operating pressure at which the calculated Kr for the given crack equals to the Kr
calculated from the above FAD equation. Therefore, a difference in KI due to the use of linear
instead of cubic interpolation method for estimating influence coefficients is carried in Kr
calculation and there on to the limiting pressure prediction.

In this example, we choose a 914.4 mm (36 in.) X65 pipe with the wall thickness of 6.35 mm (0.25
in.) and 12.7 mm (0.50 in.). The internal pressure for this example is assumed as 4,356 kPa (70%
SMYS). The cracks used in this example are defined as a/t varying from 0.1 to 0.8 and as a/c from 2
to 0.03125. Three material toughness values (20, 100 and 200 MPa√mm), Kmat are used in this
example.

Figure 9 shows the percent difference in Level 2 limiting pressure versus the percent difference in
Kr. The variation in Kr due to the use of interpolation method introduces a variation in limiting
pressure for various crack dimensions. We can infer the following from the figure –

 The difference in limiting pressure is inversely varying with the difference in Kr.
 Difference in predicted limiting pressure diminishes with higher material toughness value.
For a higher material toughness, the difference in limiting pressure is considerably lower
than a lower material toughness value. It is also observed that with a higher toughness (i.e.
Kmat = 200 MPa√mm) and shallower crack (i.e. a/t <0.4) the difference in limiting pressure
is zero or negligible.

-9-
  The maximum difference in limiting pressure observed is less than 6%. However, it can be
critical for a particular application. Therefore for a deep surface crack with low material
toughness it is important to be cautious in predicting Level 2 limiting pressure.

3%
36" X65
Wall thickness 0.25" and 0.5"
2% 70% SMYS
a/t = 0.1 to 0.8
a/c = values from Table C.12
1%
% difference in Level 2 Limiting Pressure

0%
‐4% ‐2% 0% 2% 4% 6%

‐1%
Kmat = 20 MPa√mm
Kmat = 100 MPa√mm
‐2%
Kmat = 200 MPa√mm

‐3%

‐4%

‐5%

‐6%
% difference in Kr

Figure 9 – Percent difference in the Level 2 limiting pressures

3 Conclusions
Even though the values of the influence coefficients given in the tables in API 579 fit with some
cubic polynomial functions for a/t and also for Ri/t, the nonlinearity is found more in the first
influence coefficient (Go). For a/t equal or less than 0.4, the influence coefficients can be assumed
linear. It is also concluded that shallower cracks (a/t <0.4) with a relatively high material toughness,
the linear interpolation does not produce significant difference in predicting limiting pressure with
respect to nonlinear interpolation. While for the deep surface cracks with low material toughness
may be critical for a particular application and should require caution in predicting Level 2 limiting
pressure. However the overall difference in KI as well as in Level 2 limiting pressure by using the
linear interpolation rather than the cubic polynomial interpolation is less than 6% which is
reasonably within the acceptable error margin.

4 Reference
[1] Ted L. Anderson, et. al, Development of Stress Intensity Factor Solutions for Surface and
Embedded Cracks in API 579, Bulletin 471, Welding Research Council, Inc. New York, NY, May
2002

[2] API 579-1/ASME FFS-1 2007 Fitness-For-Service, Equation C.175 and C.176.

-10-