PIPE CORROSION-CONCLUSION NEW GUIDELINES PROMISE MORE ACCURATE DAMAGE ASSESSMENT

April 16, 1990
K.E.W. Coulson, R.G. Worthingham NOVA Corp. Alberta Calgary Results of research by NOVA Corp. of Alberta have significant implications for line-pipe corrosion assessment. The background of this research and types of corrosion studied were covered in Part 1 (OGJ, Apr. 9, p. 54). Burst-test data from NOVA's research have led to more accurate damage-assessment guidelines which, if employed, could prevent unnecessary and expensive pipe repair or removals.

K.E.W. Coulson, R.G. Worthingham
NOVA Corp.
Alberta Calgary

Results of research by NOVA Corp. of Alberta have significant implications for line-pipe corrosion assessment. The background of this research and types of corrosion studied were covered in Part 1 (OGJ, Apr. 9, p. 54).

Burst-test data from NOVA's research have led to more accurate damage-assessment guidelines which, if employed, could prevent unnecessary and expensive pipe repair or removals.

BURST-TEST PROGRAM

The objective of the burst-test program was to verify the defect-acceptance criteria for the following different types of flaws: longer than 1 D, spirally oriented, and interacting.

These criteria were subsequently used to analyze data from an in-line inspection of NOVA's 500-km (310-mile), 20-in. Peace River main line system in Alberta. The program consisted of 13 burst tests performed by NOVA's works department in Spruce Grove, Alta. Fig. 1 shows a burst test for a spiral failure.

The tests were conducted on 20-in., 6.35-mm (0.250-in.) W.T., Grade 414 pipe which contained specially manufactured rectangular corrosion defects. Tests 1-12 were conducted to establish the actual failure pressure of corroded pipe with a specific flaw configuration (Table 1 and Fig. 2). The failure pressure for each test of corroded pipe was designated P.

Test 13 was performed on sound pipe with no defects to determine the expected failure pressure of undamaged pipe. The failure pressure attained from Test 13 was designated P*.

Throughout the rest of this presentation, the ratio P/P* is used to report the findings of the burst test. This ratio was utilized to provide a less complex method of reporting the results.

The defects utilized in the burst-test program were all depths of 40% W.T. This depth of penetration was chosen because it represented significant pipe defects and was also exactly half the acceptable depth of defects allowed by B31G and CSA (both codes requiring the pipe to be removed if 80% W.T. or greater has been penetrated by corrosion).

The corrosion defects were all 25.4 mm (1 in.) wide. This ensured that the flaws behaved as corrosion defects and did not fall into the controversial "narrow" defect size which, because of stress-concentration effects, could behave in a similar fashion as a crack.

This standardization of defect width and penetration was chosen to ensure that these dimensions had a constant influence on the defect performance, thereby not masking the real effect of the parameters being studied.

BURST-TEST RESULTS

The results of the burst tests (Table 2) can be summarized as follows:

  • Spiral defects are less severe than axial defects with the same projected length.

  • The failure pressure of spiral defects depends upon the length, depth, and spiral angle of the defect.

  • Circumferential and axial interaction of defects is very complex and depends on their relative length, width, depth, and separation. Flaws separated by more than the length of the shortest flaw in the axial direction are not expected to interact.

    Also, flaws separated by more than the width of the narrowest flaw in the circumferential direction are not expected to interact.

  • Little difference in failure pressure was observed for defect lengths greater than 0.75D.

  • The NG-18 surface-flaw equations were consistently conservative by approximately 2,000 kPa (290 psi). This contradicts the previous burst tests by Battelle 10 which showed the accuracy of the equations.

Recent burst testing on modern steel pipe (1980s vintage) by Battelle (references at end of Part 1) has also shown an effect similar to the conservative behavior observed in the NOVA burst tests.

SPIRAL DEFECTS

Burst testing showed the failure pressure of spiral flaws depends on axial flaw length and spiral angle (as well as depth, which was obvious and not, therefore, tested in this work).

The relationship of the flaw angle to failure pressure was important, with defects parallel to the pipe axis (90 from circumference) being more significant than those at right angles to the pipe axis.

Fig. 3 can be used in calculating the remaining strength of pipe which contains spiral flaws.

This graph of spiral angle factor (SAF) (failure pressure of an inclined flaw divided by failure pressure of an axial flaw with the same depth and axial length) vs. flaw angle shows that the original concept of using projected axial length does not apply to the case of narrow spiral flaws.

By multiplication of the calculated failure pressure of the equivalent axial defect by the SAF, a better estimate of the expected failure pressure for the spiral defect could be obtained. For example, a defect with a spiral angle of 30 (SAF of 1.2) will fail at 1.2 times the failure pressure of an equivalent axial defect.

Additional work is suggested to confirm the consistency of the SAF with defect depth and width.

CIRCUMFERENTIAL INTERACTION

The burst-test program examined two sets of flaws separated by 2 x W.T. (0.5 defect widths) and 4 x W.T (1.0 defect width).

In comparison of these two sets of flaws (Tests 11 and 12 in Fig. 2) with a single flaw of the same length (Test 5), reduction in failure pressure occurred for the samples with the multiple flaws.

Reductions in failure pressure, when compared to an identical single defect, ranged from 700 kPa (101 psi) to 200 kPa (29 psi), respectively. This showed that some level of interaction was occurring.

The amount of interaction beyond one defect-width separation, however, is expected to be minimal.

Spiral flaws are known to occur at both sides of the tape width when there is no tape overlap.

The perpendicular separation between the angled flaws corresponds to the tape width.

On the Peace River main line, this separation is 306 mm (12 in.).

In a test designed to simulate a pair of inclined flaws separated by 306 mm, defect interaction was not observed.

As this was the smallest tape width utilized on large-diameter pipelines (18-in. or 460-mm wide tape was normally used), circumferential interaction of spiral defects is not expected.

AXIAL INTERACTION, LONG DEFECTS

The burst-test program used three tests (Tests 7, 8, and 9) to determine the effect of axial spacing on flaw interaction.

The spacing between the flaws was such that, assuming the current methods of determining defect interaction were correct, the first case would definitely interact, the second would marginally interact, and the third would definitely not interact (i.e., interaction occurs with separations less than one defect length).

Fig. 4 is a graph of the expected behavior of failure pressure vs. flaw-separation distance. From Table 2, it can be seen that the three cases did not, in fact, interact.

The critical-defect interaction separation for the specific defects tested was less than the minimum separation used (0.5 defect lengths). As a result, a more detailed analysis was undertaken with the NG-18 surface-flaw equations to determine the controlling factor for flaw interaction.

The analysis showed this phenomenon to be complex.

The separation between defects at which they begin to interact varied depending upon the relative defect depths and lengths. Fig. 5 shows the interaction separations calculated for pairs of defects of the same depth.

Calculations were performed for pair depths of 5%, 40%, and 95% pipewall penetration. The interaction separations were found to range from 0.02 to 0.59 times the length of the shortest defect.

When the flaws have different depths, however, the effect illustrated in Fig. 5 becomes extremely complex. The expected interaction distances for these situations were therefore not reviewed in this program. Determining if the flaws are indeed interacting requires calculating the expected failure pressure for each flaw in the collection of flaws being assessed.

For example, consider Flaws A and B of equal length (L), separated by a distance equal to L. The calculated failure pressure for Flaw A is Pa; for Flaw B, Pb.

If A and B are interacting, the corresponding flaw would have a length of 3L, an average depth of (Da x L + Db x L + 0 x L)/3L, and a calculated failure pressure of Pab.

If Pab is less than both Pa and Pb, the collection of flaws is interacting and has a lower failure pressure than either of the individual component flaws.

If Pab is greater than either Pa or Pb, the collection of flaws is not interacting and the component flaws behave as individuals.

Little change in failure pressure was observed in the burst tests for flaws longer than 0.75D, even though NG-18 predicted longer defects to be more severe. This suggests either little real change in defect severity occurs with lengthening flaws or the change in severity is less than the variability expected in burst testing.

EFFECTS OBSERVED

The NG-18 surface-flaw equation predictions were conservative compared to the actual failure pressures observed by at least 1,869 kPa (271 psi), with a mean difference of 2,860 kPa (415 psi; Fig. 6).

This was unexpected because the equations have performed well in the past. It is possible that modern pipe steels are more resistant to crack initiation or failure, even though their strength is the same as the older steels.

In recent testing by Battelle, 10 modern steels were found to fail at pressures (stress levels) greater than what the models developed for earlier steels predicted.

Further Study is required to determine if this observed effect is real and, if so, if different "failure" criteria are required foe modern steels when defect-assessment methods are developed.

GUIDELINES

As a result of this research program, guidelines have been suggested for the assessment of corrosion-damage severity.

These guidelines, which have been used successfully by NOVA to analyze data from an in-line inspection of the 20-in. Peace River main line, are as follows:

  • The effective length of a flaw (including spiral flaws) will be the length of its projection on the axis of the pipe (axial projected length; Fig. 7). Spiral flaws will be further assessed with the spiral angle factor.

  • Flaws may interact in the axial direction if the separation between them is less than or equal to the length of the shortest flaw (not applicable to spiral flaws; Fig. 7).

  • Flaws may interact in the circumferential direction if the separation between them is less than or equal to the width of the narrowest flaw (not applicable to spiral flaws; Fig. 7).

  • Spiral flaws may interact if the separation between them along the spiral direction is less than or equal to the length (in spiral direction) of the shortest flaw (Fig. 8).

  • Spiral flaws separated by at least 300 mm (12 in.) normal to the spiral direction are not expected to interact.

  • In assessment of interacting flaws, assessment of the individual components is also necessary. The lowest calculated failure pressure of the individual components and the overall combined defect represents the expected behavior of the interacting collection of flaws.

  • Failure pressures for spiral flaws are determined, first, by calculation of the failure pressure of the equivalent axial defect; second, by determination of the spiral angle factor (SAF) from the graph in Fig. 3; and third, by multiplication of the failure pressure by the SAF.

The resulting product is the expected failure pressure for the spiral flaw.

FINDINGS; THE FUTURE

The interaction behavior of adjacent defects is not as straightforward as originally expected.

The size of the sphere of influence around two adjacent defects varies as the relative lengths, depths, widths, and separation of the two adjacent defects vary.

The work performed, however, suggests that the sphere of influence is not larger than the distance equal to the width of the narrowest defect, when circumferential interaction is considered, or equal to the length of the shortest defect for axial interaction.

Spiral defects are less severe than axial defects of the same projected axial length. The steeper the angle from the pipe axis, the less severe the defect becomes. The failure pressure of spiral defects can be estimated when the defect orientation is taken into account.

As identified in previous sections here, there are several items which require work either to confirm or better determine the expected flaw behavior.

Future work is therefore needed by the pipeline industry to complete our understanding of the behavior of corrosion defects and thereby eliminate excessively conservative safety factors in routine assessment of pipeline-corrosion damage.

Copyright 1990 Oil & Gas Journal. All Rights Reserved.