EVALUATING CORRODED PIPE-CONCLUSION PRESSURE CALCULATION FOR CORRODED PIPE DEVELOPED
Thomas J. O'Grady II
ARCO Alaska Inc.
Daniel T. Hisey
ARCO Transportation Alaska Inc.
Anchorage
John F. Kiefner
Kiefner & Associates Inc.
Columbus, Ohio
ARCO Alaska Inc. has developed a method for determining the maximum allowable operating pressure (MAOP) in a corroded pipeline. This method incorporates modifications of the ASME/ANSI B31G criterion to correct for its being overly conservative.
The method is based on procedures-outlined in Part I of this series (OGJ, Oct. 12, p. 77)-which allow evaluation of large areas of metal loss, corrosion in some welds, groups of noncontinuous pits and, to a limited extent, the effect of circumferential corrosion.
EXCESSIVELY CONSERVATIVE
The B31G criterion is based on the original Battelle work which presents a simplified set of equations for determining the effect of a corrosion defect on the hoop-stress carrying capacity of a pipe.
We will not attempt to detail the development or application of the existing B31G criterion. Both are well documented and widely understood.
Rather, we address the practical application of the modified criteria.1
The excessive conservatism embodied in the original B31G criterion may cause too much serviceable pipe to be removed. The sources of this excessive conservatism are the following:
- The expression for flow stress
- The approximation used for the Folias factor
- The parabolic representation of the metal loss (as used within B31G limits), primarily the limitation when applied to long areas of corrosion
- The inability to consider the strengthening effect of islands of full thickness or near-full thickness pipe at the ends of or between arrays of corrosion pits.
Development of the modified criterion begins as the original criterion begins, namely, with Equation 1 (see accompanying equations box).
It was known even when the original B31G criterion was developed that 110% of the specified minimum yield strength (SMYS) substantially underestimates the flow stress of a line pipe material.
For the modified criterion, therefore, the value of flow stress is taken as SMYS + 10,000 psi.
The two-term approximation for the Folias factor used in the original B31C criterion has been replaced by a more exact and less conservative approximation Of MT, as follows:
- For values of (L/Dt) 50, Equation 2 applies. L = axial or longitudinal extent of corroded area; D = OD; t = W. T.
- For values of (L2 /Dt) 50, Equation 2a applies.
The exact area of metal loss (A) is difficult to represent in terms of simple geometric shapes defined by maximum length (L) and depth (d). Two shapes considered in the development of the original B31G criterion were the rectangle (A = Ld) and the parabola (A = 2/3Ld).
Kiefner and Vieth reported on 47 burst tests of corroded pipe, the results of which indicated that the parabolic method was preferable.1
In reality, however, the parabolic method has significant limitations. Obviously, if the corroded area is very long, the effect of the metal loss is underestimated and the remaining strength would be overestimated.
To prevent misuse of the criterion in cases where long, deep corroded areas might actually have lower strengths than the criterion would predict, the method was limited to defects in which L2/Dt < 20.
This forces all long areas to be considered on the basis of 1 - d/t with a 10% allowance for flow stress.
More detailed measurements of the pit-depth profile yield more realistic representations of the metal loss. For example, a contour map of pit depths is shown in Fig. 1. Then, the "exact" profile results from the plot points along the "river bottom" path of the contour map being connected.
Calculations of remaining strength may be made on the basis of the area of missing metal in the composite profile. The depths of metal loss at regular intervals are recorded as often as every 0.25 in. The area is calculated from the resulting trapezoids as shown in Equation 3.
In that equation, do and dn ideally should be zero to indicate a return to full wall thickness. This may not always be the case, however. If the end values do and dn are zero, then Equation 4 applies.
Thus, a rectangle that is the product of the total length (Ltotal) multiplied by the average depth (davg) represents the exact area.
With the parameters Ltotal and davg thus defined, one can use Equation 2 to calculate MT based upon Ltotal. Note that A/Ao) in Equation 1 becomes davg/t.
A second method (more accurate than the parabolic method) of analyzing the remaining strength on the basis of the profile shown in Fig. 1 is called the "equivalent length" method.
In this method, the metal-loss area (A) is defined by Equation 5 and is identical to that of the "total length" method, but Leq is used as the length instead of Ltotal (Equation 6).
In this case, the area of the defect is being represented by a rectangle Leq x d. Leq is used in Equation 2 to calculate MT, and Equation 1 is used to calculate Sf. Note that A/Ao in Equation 1 becomes d/t.
THIRD METHOD
For irregular defects such as corrosion, calculating remaining strength on the basis of the total area and total length of the defect does not always lead to the minimum value of remaining strength.
The minimum remaining strength can be found by trial and error on the basis of portions of the whole defect.
A third more accurate method to predict the remaining strength involves calculations based upon various subsections of the total area of metal loss. For example, one could calculate 10 different predicted failure pressures based upon the profile shown in Fig. 2.
Each calculation involves the length Li with i varying in value between 1 and 10. The area of each individual flaw is calculated as the sum of the areas of the trapezoids made up by the discrete depth points within Li.
The procedure usually, though not always, results in a minimum predicted failure stress that is less than the value corresponding to the exact-area, total-length method.
This method, referred to as the "effective area" method, is based on the effective area and effective length of the defect. Analysis of corroded areas via the effective-area method is facilitated by the PC-program RSTRENG mentioned in the first article in this series and available from the AGA.1
Developers of the modified B31G criterion recognized that the parabolic-area representation failed to address some parameters and that some pipeline operators might not choose to use the effective-area method because of the detailed measurements required.
Therefore, in addition to the RSTRENG option, a new area representation relying solely on maximum length (L) and maximum depth of pitting (d) was conceived. The new representation is described in Equation 7; Fig. 3 shows a comparison of this new area with the parabolic.
The new A can be used in Equation 2 with the modified Folias factor, Equations 2 or 2a, to calculate predicted failure pressures when only d and L are known.
THE MODIFIED CRITERION
Arriving at a format for determining allowable values of L requires some substitutions and transformations of Equations 1 and 2.
First, let Sf = SMYS and SMYS + 10,000.
If we then define a flow stress ratio (q) with Equation 8, we can solve Equation 2 for MT (Equation 9).
Next, set the right side of Equation 2 equal to the right side of Equation 9 and solve for L. The fourth order equation yields four roots, of which only one is valid for L being sought, namely Equation 10 for values of L < 50Dt.
For values of L 50Dt, use Equation 10a.
Allowable lengths of corrosion are those which are < L, as defined in Equation 10 or 10a.
In many cases it is appropriate to use the geometric-shape approximation in which A = 0.85dL and A/Ao, = 0.85d/t. The choice of 0.85dL for A is arbitrary as was the choice of 2/3dL in the original criterion.
Without the other changes (that is, to flow stress and to the Folias factor), this new area representation would be more conservative than the parabolic representation.
The 0.85dL value becomes a reasonable choice, however, because it has wider applicability than the parabolic representation. In particular it can be safely used to analyze much longer flaws than the parabolic representation.
As a result the restrictive limits associated with the existing B31G criterion are extended significantly. These limits require reverting to a Barlow's hoop-stress calculation based on remaining thickness to determine a reduced pressure level.
As in the original B31G criterion, the cutoff value of d/t at the high end remains 0.8. Defects deeper than 80%t should be repaired or removed to prevent leakage.
The original MAOP of the pipeline is found from Equation 11 in which F = the design factor.
Under the new criterion for L < 50Dt, Equation 12 calculates a reduced operating pressure; for L 50Dt, Equation 12a.
The safe operating pressure (P') for values of L which are too large to satisfy Equation 12 are calculated with Equation 12a. In using Equations 12 and 12a, please note that P' must be < P.
CORROSION ALLOWANCE
One topic which neither the PR 3-805 project nor the resulting RSTRENG program addresses is the treatment of future corrosion.
In reviewing different methods for incorporating such a corrosion allowance in the results, we have chosen one which is neither the most conservative, nor the least. We have decided to add to the calculations the desired allowance to the measured depth of existing corrosion.
This approach results in a new maximum allowable operating pressure (MAOP) based on additional corrosion in the same location as the existing damage. We first define a P., the MAOP of the original pipe after the corrosion allowance has been taken into account (Equations 13 and 14).
Equation 15 yields a value for L < 50-Dt; Equation 15a, for L 50Dt.
The safe operating pressure, P', for values of L which are too large to satisfy Equation 15, are calculated with Equation 15a. Note that P' must be < Po.
TYPE 3 FLAWS
Type 3 flaws are those in which one of more shorter, deeper defects exist within a longer, shallower flaw. (Part 1 explains the three types of corrosion flaws.) Any one of three methods can be used for the analysis of Type 3 flaws.
First, one can always use detailed profiles of pit depths or remaining wall thickness with RSTRENG to calculate the minimum failure pressure of the "effective" length.
Second, one can always use the overall length (L) and maximum depth (d) to calculate a safe working pressure with Equations 15 and 15a. This approach, however, usually results in an overly conservative safe operating pressure (that is, well below what the pipe can actually support).
When KSTRENG is unavailable and when a reasonably accurate answer is desired, the following discussion presents a method that can be used.
Type 3 flaws may exist in the situation illustrated by Fig. 4.
A somewhat simplified approach to evaluate Type 3 defects involves the application of the B31G criterion for short defects by assuming the remaining wall thickness in the long shallow defect to be essentially uncorroded pipe of lesser wall thickness. This use of "net" remaining wall allows application of the more liberal short-defect equations in B31G.
To use this method for a shorter, deeper flaw within a longer, shallower flaw, one should first determine the lengths L1, L2, etc., and depths d1, d2, etc., of all of the shorter, deeper flaws.
The inspector should examine the pipe for a distance L/2 on both sides of these flaws to establish the minimum remaining thicknesses (tr1, tr2, etc.) that represent the deepest parts of the longer, shallower flaw,
The inspector must also determine tnet1, tnet2, etc., the actual net thickness below each of the shorter, deeper flaws.
Once this is done, the safe operating pressure levels for the shorter, deeper flaws can be determined with Equations 15 and 15a using d1, d2, etc.; L1, L2, etc.; with (and this is critical) tr1, tr2, etc. in place of t.
The "20% rule" refers to the fact that under the modified criterion, if the deepest penetration of corrosion is < 20% of the design (nominal) wall thickness (tnom) and if at least 80% of tnom remains, the segment containing such metal loss shall be considered acceptable for continued service.
Fig. 5 illustrates the application of this rule.
Measurement of the actual wall thickness (ta) is required and can be done by means of an ultrasonic device. If ta tnom required by the design, then as long as tnet 0.8nom where d = the measured depth with reference to the uncorroded surface, the corroded area may be any length.
In this case, d may be more than 20% of tnom and unlimited length would still be permitted, as shown in Fig. 5.
If ta < tnom, as in the bottom rendering of Fig. 5, the length of the corroded area (L) can be unlimited as long as tnet 0.8tnom.
Again, d is the measured depth with reference to the uncorroded surface. In this case for an unlimited length to be permitted, d must always be < 20% of tnom.
Note that d is not compared to tal, the actual measured thickness, in application of the 20% rule. The depth d is always compared to tnom, and the criterion tnet 0.8tnom must always be met for L to be unlimited.
Note also that the 20% rule does not allow portions of corroded areas to be neglected in the determination of L. Specifically, when a tnet of < 0.8tnom is encountered, the 20% rule must not be applied to those portions where tnet is not < 0.8tnom- All of the corroded area is to be considered in determining L. Only if the effective length as determined by RSTRENG is < L is one permitted to consider a length < L.
CIRCUMFERENTIAL CORROSION
Only rarely does one need to be concerned with the circumferential extent of corrosion in a pressurized pipeline. Such concerns exist only when one of three conditions exist:
- The circumferential extent is large in comparison to the longitudinal extent, and the longitudinal extent is too small to cause the pipeline to be repaired or the repair required will not correct the circumferential weakness; or,
- The pipeline is subjected to unusually high longitudinal tensile stress; or,
- The corrosion exists in an area of the compressive stress, and the metal loss will endanger the pipeline from the standpoint of local buckling.
The following shows methods for assessing the effects of the circumferential corrosion. These methods, although unsophisticated and therefore embodying large safety factors, should nonetheless provide adequate means of protecting the pipelines from a corrosion-caused circumferential mode of failure.
In Fig. 5, one notes that the length of a circumferentially oriented part-through flaw is considered in terms of the fraction of the circumference affected (either in terms of an angle, 2a, or an arc length of C = 2Ro.). Depth of the flaw is expressed in terms of an assumed uniform depth (d) or the d/t ratio.
In the discussion of circumferential defects, the symbol SL will also indicate that the concern is with longitudinal stress instead of hoop stress as in the case of longitudinally oriented flaws.
Whether the flaw is internal or external makes no difference in the analytical method to be considered. Here we have shown the external flaw embodiment.
The analytical technique presented considers the flaw for analytical purposes to be the area C multiplied by d. No accounting of the exact area of an irregular flaw is made; all are characterized by maximum length and maximum depth.
Furthermore, the technique is based on the assumption that the flaw is symmetrical about an axis perpendicular to the neutral axis of the flawed cross section.
The flaw is assumed to be positioned either on the tensile side or the compressive side of the neutral axis as far from the neutral axis as possible so that it will be subjected to the maximum tensile or compressive stress for any combination of bending or axial load.
A circumferential flaw is affected primarily by longitudinal stress in the pipe and not by hoop stress. Thus, knowledge of the longitudinal stress is essential to the analysis of such flaws.
Longitudinal stresses in pipelines and piping systems arise from both bending and axial loads.
Bending loads generally arise from external sources, although internal pressure itself causes some bending stress on a flawed cross section because the center of application of the pressure force does not coincide with the neutral axis. (Note that the neutral axis in Fig. 6 is displaced from the center of pipe.)
Axial loads may arise from a number of sources including pressure, restraint of thermal contraction or expansion, and external loads.
One major difference between the analysis of longitudinal flaws and circumferential flaws is the degree of complexity of the stress situation. For longitudinal flaws, the hoop stress arising from pressure is usually the only stress-producing factor.
In complex piping systems either above or below ground, it is often difficult to determine the true state of longitudinal stress. In addition, the consideration of uniform axial stress (as opposed to bending stress) causes difficulties in any attempt at a simplified analysis.
Therefore, in this method the uniform longitudinal stress will be treated in a severely simplified manner. The recommended factor of safety will more than compensate for this simplified analysis.
WILKOWSKI-EIBER METHOD
Several years ago in response to a need in the offshore pipeline industry, Wilkowski and Eiber developed a method for predicting the critical size of a circumferential repair groove that could be made at a girth weld in the pipeline as it was being laid into water from a pipelaying barge.2
The approach was similar in principle to that used in the analysis of longitudinal flaws described previously here.
An empirical stress-intensification factor (Mc) analogous to the Folias factor was derived from test data. The factor, which is greater than unity, is a function of the through-wall flaw's arc length and is divided into the-flow stress of the material (S) to obtain the predicted longitudinal bending stress that will produce failure of a through-wall flaw.
The failure stress levels for surface flaws can then be calculated on the basis of this empirical through-wall flaw intensification factor.
The Wilkowski-Eiber method2 was intended for application to relatively narrow defects (repair grooves) in pipes subjected to bending stress only (no uniform axial stress).
As such, any possible toughness dependency is ignored. For evaluating the effects of the circumferential extent of the corrosion in an area of tensile stress, however, this method may be useful.
The method can best be illustrated by means of sample calculations. For an example, we will use a 30-in. OD, 0.344-in. W.T. X60 material with an assumed flow stress of 70,000 psi.
The defects which will be used in the example are five different lengths of circumferentially oriented surface flaws with d/t values of 0.5, 0.6, 0.7, and 0.8. Referring to Fig. 6, one may note that the key parameters of the five lengths of defects are as shown in Table 1.
For this method, one needs only the flow stress (;) and the flaw and pipe geometries to calculate the failure stress with Equations 16 and 17.
The results are shown in Table 2 (recall that d/t varies from 0.5 to 0.8.).
According to ASME B31.4, the sum of stresses from "pressure, weight, and other sustained loadings shall not exceed 75 percent of the allowable stress" (SA) which is 72% of SMYS.3 Thus, the maximum expected longitudinal stress is 54% of SMYS or 32,400 psi.
If a factor of safety against failure of a circumferential flaw is taken as about 1.5, it would appear that circumferential corrosion centered on the axis perpendicular to the neutral axis of the pipe can safely be left unrepaired if it is:
- Less than half way through the wall thickness (< t); or,
- More than half way through the wall thickness but less than 0.6 through the wall and < 1/6 of the circumference in extent ( 0.5t but < 0.6t and < 0.17C); or,
- More than 0.6 through the wall but < 0.8 through the wall and < 1/12 of the circumference ( 0.6t but 0.8t and < 0.08C).
On this basis, Part 1 provided guidelines for judging the need for repair of corroded pipe based on the circumferential extent of corrosion.
As significant amounts of metal are lost to internal corrosion, local buckling becomes a possibility.
For weight loads, the weight of the pipe and its contents result in a uniformly distributed load along the pipe. These loads induce a concentrated vertical reaction and a bending moment in the pipe at the support.
If the wall is excessively thinned, these concentrated forces may be large enough to cause severe deformation of the pipe at the supports. The bending moment may cause yielding and buckling of a thin-wall section.
A conservative estimate of the compressive bending stress and strain can be obtained by assuming that the wall of the pipe is reduced uniformly to thickness t, at the support but remains equal to its original thickness (t) everywhere else.
The uniformly distributed weight of the pipe (w) is, therefore, approximated by Equation 18.
The maximum resulting bending moment is estimated by assuming the supports are evenly spaced at the maximum actual spacing. The moment (m) is thus given by the "fixed-end moment formula," Equation 19.
The bending stress (Sb) at the support with reduced wall (t,) is approximated in Equation 20.
If this stress exceeds the pipe yield stress, remedial action would be necessary to prevent excessive deformation. As a further check on bending, the strain at buckling (Ebk) is estimated by Equation 21; the bending strain (Eb), by Equation 22.
Therefore, because the bending strain is less than that required for buckling (Ebk), local buckling is unlikely. Thus, as in the case of the effect of tensile longitudinal stress, loss of one-half the wall thickness is generally not expected to cause problems in areas of compressive stress.
REFERENCES
- Kiefner, J. F., and Vieth, P. H., "A Modified Criterion for Evaluating the Remaining Strength of Corroded Pipe," Project PR 3-805: Pipeline Search Committee, American Gas Association, Dec. 22, 1989.
- Wilkowski, G. M., and Eiber, R. J., "Determination of the Maximum Size of Girth Weld Repair on Offshore Pipelines," American Gas Association, 1979.
- Paragraph 402.3.2.(d), ASME B31.4, 1989 Edition, Liquid Transportation Systems for Hydrocarbons, Liquid Petroleum Gas, Anhydrous Ammonia and Alcohols.
Copyright 1992 Oil & Gas Journal. All Rights Reserved.