Proceedings of the Eleventh (2001) International Offshore and Polar Engineering Conference Stavanger, Norway, June 17-22, 2001

Copyright 2001 by The International Society of Offshore and Polar Engineers ISBN 1-880653-51-6 (Set); ISBN 1-880653-53-2 (VoL I1); ISSN 1098-6189 (Set)

Background and Development of DNV-RP-F101 "Corroded Pipelines"

O.H. Bjornoy, G. Sigurdsson and M.J. Marley
Det Norske Veritas Oslo, Norway

ABSTRACT The DNV Recommended Practice for the assessment of corroded pipelines, DNV-RP-F101 was issued in 1999, and has been used with success in many projects. In this paper the development of the RP is outlined, the determination of the safety factors are explained and the sensitivity of the safety level to the various parameters is demonstrated. RP-F101 includes recommendations for the assessment of: single defects; interaction of defects; complex shaped defects; and additional external loading. RP-F101 describes two alternative approaches with different safety philosophy. The first approach includes calibrated safety factors, taking into account the natural spread in material properties and wall thickness and internal pressure variations. Uncertainties associated with the sizing of the defect and the specification of the material properties are specifically considered in determination of the allowable operating pressure. The second approach is based on the allowable stress design format, where the allowable operating pressure is determined from the capacity and multiplied with a single usage factor based on the original design factor. The development of the capacity equations and acceptance equations giving the allowable operating pressure and the development of the safety factors are described in this paper. The RP was a result of a co-operation between BG Technology and DNV. The results from their respective joint industry projects have been merged and form the technical basis for RP-F101. Several oil companies and regulating authorities have sponsored the development of the RP. KEY WORDS: pipeline, burst, reliability, corrosion, calibration INTRODUCTION The development of a new code for the assessment of corroded pipelines was a major task, comprising a number of steps and conducted over a period of several years. This included evaluation of existing methods and test data, investigation of the failure mechanisms for burst and leakage, parametric analyses using FE methods, full-scale pressure tests to failure, development of capacity equations based on analyses and test results, defining safety philosophy and development of acceptance equations including

safety factors for the determination of the acceptable operating pressure with a defined safety (reliability) level. The steps in the development and the background for the methods are described in this paper DNV was running a JIP project looking into the available methods for the assessment of defects in pipelines, and identified the need and potential to develop improved methods tbr the assessment of corrosion defects in pipelines including safety thctors calibrated by use of reliability. The DNV project (Bjorne~y 1997) included 12 full-scale burst tests with machined corrosion defects and a development of a comprehensive database of 3D non-linear finite element analyses of pipes containing defects. Safety factors were determined using reliability methods. At the same time, BG Technology was running a similar project (Fu 1995, Batte 1997, Fu 1998) where more than 70 burst tests of pipes containing machined corrosion defects were performed, including single defects, interacting defects and complex shaped defects. In addition, a comprehensive database of 3D non-linear FE analyses of pipes containing defects and a database of linepipe material properties were generated. BG Technology and DNV merged the technical results from the two respective projects and developed a unified guideline for assessment of corrosion in pipelines. The unified guideline has formed the basis for the DNV-RP-F101 "Corroded Pipelines", (Sigurdsson 1999, Bjornoy 1999) The safety factors are calibrated for 3 different reliability levels, associated with safety class target reliability levels as defined in DNV-OS-F101, Submarine Pipeline Systems. The target reliability levels are based on the offshore pipeline standard, but DNV-RP-F101 is not limited to the assessment of offshore pipelines; it may be applied to pipelines, onshore and offshore, or to piping designed to other codes, as long as the safety philosophy in the codes is not violated. DNV-RP-F101 is not a replacement of the widely used ASME B31G code, but an addition when a plastic failure collapse can be expected. Modern pipeline steel materials have normally sufficient toughness to expect plastic collapse failure. Battelle (Stephens and Leis, 2000) has developed a capacity equation with similar results, and have given a preliminary recommendation of a Charpy value of 45 flabs (about 60 J) for the lower bound material toughness for plastic collapse. DNV-RP-F101 has limited the application to defects where a plastic collapse are expected, but without defining 102

any lower bound for the material toughness. As the Battelle and DNV-RP-F101 give similar results, the proposed Charpy value would also be applicable for the DNV-RP-F101 code. SINGLE DEFECT In the development of the assessment methods, much effort was given to single corrosion defects exposed to internal pressure. The basis for the capacity equation and the acceptance equation is extensive, with a large number of finite element analyses where the effect of all important parameters has been investigated, and a large number of full scale burst tests performed. The single defect equation forms the basis for the assessment approach of combined intemal pressure and compressive longitudinal stresses, the interaction of single defects, and the complex shape of a corrosion defect, which also demonstrate the need to have a sound basis in the single defect equations. Capacity modelling The expression of the burst capacity for a single, longitudinally oriented, rectangular shaped, corrosion defect was developed based on a large number of FE analyses, and a series of full scale burst tests. By using finite element analyses the effect of each important parameter was investigated for a reasonable cost, while the goodness of the analyses were verified by a large number of fullscale burst tests. BG developed the failure criterion used in the FE analyses, and the FE predicted failure pressure showed very good agreement of with tests results (Fu and Kirkwood,1995). Over 400 three-dimensional FE analyses were performed to generate a database of failure pressures for variations of diameter-thickness ratio, corrosion length and depth and material properties. The burst pressure capacity is expressed as: PCAP=YlabYFEA Pfit where Pj~ : predicted capacity from FE capacity equation : bias in (Eq. 1)

Comparison between capacity Equation and FEA

. .o

0.8 0.6 0A 0.2

0= i

~J

z 0 02, 0.4 0.6 0.8 Normalised Capacity -FEA results

.... !..... i....... ! ...... ?..... !. . . . . . . . . . . i...... !............. . . ; ........ '

Figure 1 Comparison between the capacity equation Pfit and FEA results (normalised) For practical purposes, the FE capacity equation Pfit replaces all finite element analyses, with an excellent accuracy. Again, the finite element analyses were conducted to investigate the sensitivity to the important parameters for the whole validity range, which would be costly and impractical to cover in laboratory tests. The burst capacity is well predicted by using the tensile strength or,, but the yield to tensile ratio t~y/cr u also have an influence. It is observed that for very short defects, long defects and shallow defects (and plain pipe) the tensile strength is a good measure for the burst capacity. For medium length defects, the burst capacity is also sensitive to the oy/au ratio, depending on the defect length and defect depth. This is illustrated in fig. 2, The effect is normalised to a ~y/O, ratio of 0.77, and is shown for the range from 0.6, 0.7, 0.8 and 0.9, for a defect depth of d/t = 0.3 and 0.6. For d/t = 0.6, the capacity is up to 10% higher for a pipe with a high yielding level, and about 10% lower for one with a low yield level, both with the same tensile strength. This effect decreases for shallower defects and for a defect depth d/t = 0.3 the corresponding effect is about 5%, and becomes zero when the defects decreases to zero.

YFf.A

P~t compared to numerical FE analyses

Y/,,~ : bias in the FE analyses compared to laboratory tests A capacity equation Pfit was obtained through a multivariable curve fitting of the FEA results. This equation should be a best possible representation of the finite element results, and the form and complexity are less important as this equation is to be used only in the calibration of the safety factors. d Pfit

1.10

"

'

=YB~

f2 ' ~-'t'

(Eq. 2)

...........

t fl(~-Dt ] where D d outer diameter depth of corrosion defect tensile strength L = length of corrosion defect YB (an empirically derived term) accounts for the boundary conditions of the pipe (see Sigurdsson 1999). The functions J)( ) andj~( ) were obtained through the curve fitting analysis. The above fitted FE based equation predicts the finite element results extremely well, unbiased (mean = 1.0) and with a coefficient of variation (CoV) of only 2 percent (YFzA), and is shown in fig. 1. = =
",, O'y / C u = ,7

0 90

~ 6

' 8

10

Normalized length

Figure 2 Capacity dependence on yield to tensile ratio.

103

The next step in the development was to compare the FE results (not Pfit) with the corresponding laboratory burst test results. This comparison determines how well the finite element results predict the "true" capacity. In the laboratory tests all parameters are accurately measured. The FE results were compared with the full scale tests by BG Technology and DNV, and found to unbiased and with a CoV of the order of 7-8% (Ytab). The BG Technology burst test and FE results remain confidential, and a detailed graphical comparison is not shown. So far it is determined that the capacity equation P~ represents the FE results with a CoV of 2%, and that the FE results predict the laboratory burst capacity with a CoV of about 7-8%, both unbiased. Together this will result in a capacity equation (eq. 1) which is unbiased and with a CoV of 8-9%. Preferably the capacity equation should have been based directly on test results, but this would require many burst tests and should contain results for the full validity range, dimensions, material grades, defect sizes etc. The selected approach, with FE analyses for parametric sensitivity study and correlated against burst tests, is preferred when limited number of tests is available. It would be possible to develop a good fit to the appropriate burst tests, with a lower CoV than the above mentioned CoV of 8-9%. The statistics of the burst capacity prediction are essential information for the calibration of the safety factors. Much effort was made in the evaluation of the applicability of each single test, deciding if it should be included or excluded in the determination of the mean and standard deviation. It is generally favourable to include as many tests as possible, but it is crucial that the tests are representative. The capacity equation is designed to cover rectangular shaped defects, and therefore only tests with rectangular shaped defects can be used when the test results are compared to the equation. If other types of shapes are used the effect of this needs to be accounted for, for example using a shape correction factor. It is not satisfactory with nominal dimensions or material grade only, the material properties should be known. It should be mentioned that most of the published test data were excluded, due to lack of data or the shape of the defect was other than rectangular. If irregular or parabolic defect shapes are included in the comparison, and the maximum depth and length are used, the capacity equation would underestimate the failure pressure, as the defect is not as large as the rectangular shaped defect assumed in the equation. This would erroneously result in a higher mean value, which is unconservative, but most probably also result in a larger standard deviation. However, when the acceptance equation including the calibrated safety factors is defined, this equation can be used for irregular and parabolic defect shapes using the maximum depth and length. The actual defect will be enveloped by the rectangular shaped defect defined by the maximum depth and length, and hence the prediction will be conservative. Alternatively, if the defect profile is known in detail, the complex shape approach in RP-F101 can be applied. Simplified capacity equation The capacity equation as defined in eq. l and 2 is complicated and impractical in use. A more "user-friendly" and simplified capacity equation for the burst capacity is introduced and is similar to the capacity equation implemented in DNV-RP-FI01. 2t- cru (1 - ( d / t ) ) Pcap=XM ( D - t ) ( l _ ( % t ) ) (Eq. 3)

where

Q=

1 + 0.31

The XM accounts for the bias and the model uncertainty, and is described in the following. This equation is for all practical purposes sufficiently accurate for the capacity assessment of corroded pipelines. The equation can also form the basis in reliability analyses of corroded pipelines, as tong as appropriate distributions of the parameters and model uncertainty are included. It is of interest to see how well the capacity equation (eq. 3) predicts the burst capacity of published tests. The comparison are made for the BG Technology tests, tests published by Chouchaoui and Pick 1992; Coulson and Wolthingham, 1990; and Kiefner and Vieth 1994, and tests recently published by Benjamin 2000 and Bjornoy 2000 (2 tests with only internal pressure). Table I. Comparison with burst tests (tests / P~.p): XM Reference No. tests Mean CoV Chouchaoui and Pick Coulson and Wolthingham, 23 1.00 8.9% Kiefner and Vieth BG Technology 34 1.05 6.1% Petrobras and DNV 9+2 1.09 2.7% All 68 1.04 7.3% The results in the table indicate that the simplified equation 3 predicts the burst capacity as well as the complex equation (eq. 1 and 2), but the burst tests represents a minor variation in material properties, pipe dimensions and defects sizes. This correlation would have most probably have increased if the tests represented a wider variations in parameters. The simplified equation could be used instead of the more complex equation (eq. 1), and the factors Ylab, YVEAand YB would then be represented by the single factor XM. The values of the Y factors have been recalculated replacing Pnt with Pcap (without XM) and combined into XM, being normal distributed with a mean value of 1.05 and CoV of 9.5% (StD = 10%). Probabilistic assessments of pipes with corrosion defects can be based on the following limit state function: g = Pcap - PrNT (Eq. 4)

where Ptwr is the annual maximum differential pressure and Pcap is the burst pressure capacity as defined in equation 3. The parameters in the limit state should be modelled with its actual distributions, and considerations should be given to the inspection sizing accuracy. A set of distributions considered to be representative for pipelines were used in the calibration of the safety factors included in RP-FI01, and are given in by Sigurdsson 1999. Table 2. Parameters in the modelling of the burst limit state Variable Distribution Mean CoV PINT Gumbel 1.05 MAOP 3.0 % D Deterministic Actual t Normal Nominal 3.0 % ~u Normal 1.09 SMTS 3.0 % - 6.0 % lmeas Normal Measured value Specified d/t Normal Measured value Specified XM Normal 1.05 9.5 %

104

The safety factors in DNV-RP-F101 account for the inspection accuracy. Growth of the corrosion defects are not directly accounted for, but by using probabilistic methods this can easily be included and the failure probability as function of time can be calculated. The corrosion rate will also add to the uncertainty of the future defect size, in addition to the inspection accuracy. The defect depth dr and length l T a t time T may be calculated as : (Eq. 5) d T = d o + T . rcorr IT = I0 + T - rcorr, length where dT = corrosion depth at time T do measured corrosion depth T = time interval in years from last inspection rcorr = average annual corrosion rate IT = longitudinal defect length at time T l0 = measured longitudinal defect length rcorr, length = average annual corrosion rate in length direction
=

Safety philosophy
DNV-RP-F101 describes two alternative approaches for assessing corroded pipelines. The main difference in the two approaches is in their safety philosophy. The first approach is in accordance to the safety philosophy adopted in the DNV Offshore Standard OS-FI01, Submarine Pipeline Systems. Uncertainties associated with the sizing of the defect depth and the material properties are specifically considered. Probabilistic calibrated equations (with partial safety factors) for the determination of the allowable operating pressure of a corroded pipeline are developed to ensure a consistent reliability level for different designs. The second approach is based on the ASD (Allowable Stress Design) ]brmat. The failure pressure (capacity) of the corroded pipe is calculated, and this failure pressure is multiplied by a single utilisation thctor based on the original design factor. Consideration of the uncertainties associated with the sizing of the corrosion defect is left to the judgement of the user. The basic equations used for the two approaches are the same. In the first approach partial safety factors are introduced, accounting lbr level of uncertainty, while in the second approach only a single utilisation factor is used.

as suitability of the equation, material properties and sizing accuracy of measured defects. First it is assumed that the defect and the material properties can be accurately measured (as in a laboratory burst test). The next step in the example illustrates the effect of not being able to take material tests and the pressure may slightly vary as in a real pipeline. The last step illustrates the effect on the safety level when the inspection method is not able to exactly determine the detect size, and the sizing is associated with a certain level of uncertainty. Figure 3 illustrates the burst capacity (eq.3) as function of the defect length for a pipeline for a defect depth of 45 % of the wall thickness. A rectangular shaped smooth shaped defect is assumed. All parameters are known, dimensions and material properties are measured and tested. The full line is the best estimate of the capacity of a burst test (X~a = 1.05). However, even if all parameters are known to our best ability, we can not exactly predict the burst capacity, and the result for a single test can as well fall above the line as below, but most probably it will fall close to the line. If many tests with the same defect were performed, it is expected that half of the tests would fall above the line and half of the tests would fall below the line, and most of the test would fall close to the line. As mentioned earlier, the capacity equation, represented by this line, was made as a best fit to FE and burst test results. The line is the 50% failure probability line. A 406mm (16") diameter and 12.7 mm thick pipe of grade X60 (SMTS = 517 MPa) linepipe with a defect of 45% of the wall thickness is used as the example. From material testing an average or, of 579 MPa is measured.
Burst pressure for corrosion defects
400 _ _

.....
300

i- -&-;Zp;gni-~-iiiiy-(i~;-si

iJi ....

i .....

'~ 2~

i
il,

200

_---=_ --"- .-= ~ ....

'

_'-.':~_ __

100 ti ......

" .....

10"'

probability

' - .....

~ .........

i ........

Acceptance equation, calibrated safety factors (Part A)

The capacity equation predicts the burst capacity with a good accuracy, for known defect size, pipe dimensions and material properties. However, these parameters are usually not known without a certain degree of uncertainty, and this should be accounted for in calculating the acceptable operating pressure of the corroded pipeline. The FE analyses results and the laboratory tests results, together with statistical data of material properties, pressure variations and selected levels of uncertainties in the defect sizing form the required basis for a reliability code calibration. It is reasonable to allow for a less reliable operation of a pipeline with modest consequences in case of a failure, compared with a pipeline with enormous cost, safety or environmental failure consequences. To cover these aspects, the target reliability levels were selected in accordance with the safety class classification in DNV-OS-FI01. The annual target . . . . -4 failure probabilmes assocmted with these safety classes are 10 10-5 for the Normal Safety Class, 10"3-10-4 for Low Safety Class and <10 -5 for High Safety Class. In the example described in the subsequent pages the effect on the safety level from the various parameters are demonstrated, such
0
0

.......

) ........

t,o-.

......

i .........

100

200

300

400

500

600

700

Defect length ( r a m }

Figure 3 Capacity equation, known parameters.

Model uncertain~ The burst capacity equation predicts the full-scale tests with a standard deviation of 10%, see factor XM described earlier. Using this information, additional lines with selected failure probabilities can be drawn. The 1% (10 "2) failure probability line is shown, where one out of 100 tests is expected to fall below this line. And further, the 10-3 and 10-4 failure probability lines are also shown, where one out of I000 and I0000 tests is expected to fall below the lines, respectively. The lines is determined according to the formula: P = Pep (I-[3.o) where ~ is the standard deviation of XM of 10%, and 13 is the socalled reliability index e.g. q)(-13) = Pf from normal distribution, 2.32 for 10-2, 3.09 for 10-3 and 3.71 for 10"4. From figure 3, the best estimate of a burst test for a 45% deep and 300 mm long defect is 262 bar. The 1% probability is 201 bar,

105

10-3 is 181 bar and 104 is 165 bar, which is a significant reduction compared to the best estimate capacity, a factor of 0.63. The dotted line indicates a reasonable design pressure for Normal safety class of the pipeline used in this example. Uncertainties in material and pressure The actual material properties will normally not be available in much detail at the location of a corrosion defect in a pipeline as it is in a laboratory burst test, and the uncertainties in the material properties also need to be accounted for. The steel grade is usually known, and even in cases where material certificates with test results from the steel mill are available there are still uncertainties related to the material properties at the location of the defect. There will also be some additional uncertainties to the internal pressure compared to a laboratory test, as there may be some fluctuations over time, and some pressure upsets. Taking these uncertainties into account will result in a lower acceptable pressure compared to the lines given above. Figure 3 is copied in Figure 4 which also shows the further reduction in maximum allowable operation pressure due to the defect for a failure probability of 10"4. The reduction from the second lowest to the lowest line is due to the material uncertainties and pressure variations. The distributions for the material properties and pressures variations are obtained from the Superb project (Jiao, 1995). The material tensile strength % is on average given to be 9% higher than SMTS, and is categorised with a CoV of 6%.
Buret pressure for corrosion defects
400 __ : : : : : :

the measured depth, while the shade area indicated the range of the most probable depth of the defect. In this example the inspection accuracy is +15% oft, stated with a confidence level of 80%, which is a common term by the inspection contractors. This is interpreted as there is 80% probability that the actual defect depth is within +15% o f t of the measured depth. This implies that there is a 20 % possibility that the actual depth is outside this range, and if it is equally distributed, it is a 10% probability that the defect is deeper than 60% when the measured depth is 45% of t. In figure 5 this effect is illustrated with the wide distribution. A distribution with much better accuracy is included to demonstrate the benefit of an improved inspection.

Figure 5 Measured defect depth and sizing accuracy

The effect of the inspection accuracy, combined with the other uncertainties described earlier, is accounted for in the calibration of the safety factor in DNV-RP-F101. Preferably one would like to have a single safety factor for easy use, but to cover all combinations of the above effects it is much easier and less conservative to apply several partial safety factors, accounting for the uncertainties from each parameter. Initially, the calibration considered four partial safety factors, but by examining the sensitivity of each of the factors, the number of factors was reduced to two without reducing the applicability of the acceptance equation. Partial safety factors on the wall thickness and defect length were excluded, but these are accounted for and covered by the combination of the other factors. Please see DNVRP-F 101 for details of the safety factors. In DNV-RP-F101 the maximum allowable operating pressure, P,,,~,o, for a pipeline with a corrosion defect is not allowed to exceed the following pressure Pco,.r:

3s0 ! ......
300

i ....

! Jso';,~;ro-/'a~i;ffY-t-bWsiY'i)--~-! .......
.....

250

~ 200
.~ 15o

so J . . . . .
[
o

i ..... 1
:

j 10""
200

p r o b abi'llt y " . . . . . . . . .

i .........

! ........

10 "" p r o b a b
:

prob'ability

2 t SMTS (1 ,_

rd(d/t) *)

[
0

b.~,.,~t~,

Pcrr=Ym
700

(D-t)(1

100

300 400 500 Defect length (ram)

600

rO(ot)*)

(Eq. 6)

Figure 4 Acceptable operating pressure, known defect size

The reduction of operating pressure for 10-4 case is clearly illustrated in the figure 4. In case the material were fabricated with stricter tolerances and control, or the pressure control system were stricter, the reduction in allowable pressure would be less. Uncertainties in defect sizing The pressure given in figure 4 is the acceptable operating pressure for a pipeline with a corrosion defect of 45% depth, and where the defect size is "exactly" known. In some few cases this can be achieved, or at least almost, but normally the sizing of the defect is associated with some level of uncertainty. If the defect is measured to be, say 45% of t, but there is a possibility that the true size is not 45,/0, but as deep as 60%, or maybe less deep, say 30%, this of course also needs to be accounted for. The potential less deep defect does not cause much of concern, but can result in costly unnecessary repairs or operational restrictions. In this context the possibility that the defect is deeper than measured causes concern. The full line in figure 5 illustrates

where Q= 1+0.31

(d / t)* = (d / t)mea s + ~'d "StD[d / t] and D d t L

(d/t) .....

= = = =
=

?m 7d
E;d

= =
=

SO[d/t] = SMTS =

outer diameter depth of corrosion defect wall thickness (nominal) measured length of corrosion defect measured relative corrosion depth partial safety factor for model prediction partial safety factor for corrosion depth factor defining a fractile value for corrosion depth standard deviation of the measured (d/t) ratio (based on the specification of the inspection tool). Specified Minimum Tensile Strength

106

Note that the measured values of the defect depth and length are input to the acceptance equation, and that the inspection accuracy also is input. The depth to be used in the equation is the measured depth plus an additional depth depending on the inspection method and accuracy, given by (d/t)*. This value again is multiplied with a partial safety factor. Safety factors are shown below, see DNV-RPFi01 for lull details. Partial safety factor 7,, = 0.74 for safety class Normal (10 -4 annual failure probability).

accuracy is the line indicated with "Exacf'. The acceptable operating pressure becomes very sensitive to the inspection accuracy when it exceeds +10%, for a measured depth of 45%. 400 .......... ~. . . . . . s ~ o
Burst pressure for corrosion d e f e c t s
=

........ :-F ........ -7- .

. . . .

......... ~.......... ! "

~......... i ii . . .l. . . . i. ... . . .i. i ......

Table 3. Partial Safety Factor Inspection sizing Accuracy, StDld/tl

(exact) 0.00 0.04 0.08 0.16

ed
0.0 0.0 1.0 2.0

Low

Safety Class Normal "/d= 1.00

~td= 1.16 Itd = 1.28 ~'d = 1.38

High

)'d = 1.00
"t'd= 1.16 Yd= 1.20 )'d = 1.20

Td= 1.00
I'd = 1.16 )'d = 1.32 ~'d= 1.58
100 200 300 400 500 Meaeursd Defect length (ram) 600 700

Inspection accuracy The inspection method and the standard deviation of the inspection accuracy and the safety class determine the partial safety factors. For the case where the defect size is exactly known, the measured (exact) defect size is used in the equation. If the inspection accuracy is quoted with _+10% of t, stated with a confidence level of 80%, then the standard deviation is 0.08, and the factors are determined as ~d =1.0 and ya=l.28 from the table above. The measured depth is 45% oft, and the input into the equation is: ( d / t ) * = ( d / t ) m e a s + ~'d " S t D [ d / t ] (d/t)* = 0.45 + 1.0 0.08 = 0.53 This is multiplied with ya=l.28, resulting in a total of depth 0.68 d/t to be used in the equations. This illustrates the importance and sensitivity to the inspection accuracy. In the continuation of the example, the acceptance equation Pc,,,~ is used. The acceptance equation Pcorr includes predefined calibrated safety factors and is easy to use, and with exact known defect size it should give about the same pressure as the reliability based allowable pressure, the lowermost line in figure 4. However, Peo~ is calibrated using the capacity equation including P~t and will not result in exactly the same results as Pco~r,but may vary depending of the selected case, and for this case it is slightly higher. Curves from the previous figures are included for illustration and comparison. All curves are for normal safety class, with a target annual failure probability of 104. The uppermost curve is the same as presented earlier, the best estimated capacity. The second is the 10 "4 annual failure probability, including uncertainties associated with the material, but where the inspection accuracy is exact. From this line the effect of the inspection tool sizing capabilities can be illustrated. Three additional levels of inspection accuracy is shown, _+5% of t, _+10% and _+15 %, all quoted with a confidence level of 80%. It is reasonable to require reduced allowable operating pressure for a defect measured with large uncertainties, as it as well could be deeper, than an accurately sized defect measured to be the same depth. The reduction in acceptable operation pressure for a measured defect size of 45% of t is clearly illustrated in figure 6. The acceptable operating pressure is only 108 bar for a defect length of about 300mm for an accuracy of _+10%. By improving the inspection accuracy, and measuring the same size, the acceptable operating pressure will be increased accordingly, as illustrated by the second lowest line where the inspection accuracy is _+5% with 80% confidence. The maximum benefit of improving the inspection

Figure 6 Allowable operation pressure for measured defect depths for various inspection accuracy with 80% confidence level. (Measured 45% of t) Acceptance equation, ASD format (Part B) The alternative safety philosophy format is based on the ASD (Allowable Stress Design) format, where the failure pressure (capacity) of the corroded pipe is calculated, and an appropriate safety factor is applied. This factor may be based on the original design factor, but the user also has to consider the uncertainties associated with the sizing of the corrosion defect. This has been the common approach in the industry until today. It should again be highlighted that Part A includes calibrated safety factors for specified safety levels, and is in most cases the recommended method.
2t'~ u (1-(d/t)) (D_t) (l_(d~t))

Pf=

(Eq. 7)

where Q = +0.31

Psafe = 0.9.F.Pr
An appropriate safety factor for determining the safe operating pressure may be based on the original design factor and multiplied by the 0.9 factor to account for the additional uncertainties due to the presence of a corrosion defect. Again, please note that the user has to give due consideration to the measuring uncertainties when selecting the safety factors. (There is a common misunderstanding that the 0.9 factor mentioned above apply when cr, is used, but should be omitted when SMTS is used. This is not the case, the 0.9 factor apply in both cases. The reason for this misunderstanding is probably because BS 7910 (1999) has omitted this factor, resulting in a 10% higher calculated safe operating pressure compared to DNV-RPF101 for the same safety factor.) In RP-FI01 Pf is labelled as "failure pressure", but please note that the factor XM is not included, and that results in a 5% lower

107

calculated failure pressure compared to the best estimate, equation 3. Further, if au (UTS in DNV-RP-FI01) is not known, which usually is the case, ~u should be substituted by SMTS, which results in additional conservatism of about 9% (and more), resulting in a total conservative estimate of 14% (tru is estimated to 1.09 SMTS, see Superb, Jiao, 1995). The calculated Pf includes a safety margin of about 14% and more whert SMTS is used.
Burst pressure for corrosion defects
400

defects can be accepted before remedial actions are required. Further, the maximum actual corrosion defect that can exist before the integrity of the pipeline is jeopardised can be calculated by assuming exact accuracy of the inspection tool, see figure 8. Note that the y-axis of the figure now is the acceptable measured relative defect depth, show as function of the defect length, and is presented for selected combinations of operating pressure and inspection accuracy.
1.0 0.9

Burst pressure f o r corrosion defects

i
t

300

0.8 0.7

125o
~200
100 5O
0 . . . . . .

!0.o
.* o.s
0.4
b

!i'

0.2 0.1 : 100 200 300 400 500 600 i 700

100

200 300 400 500 M e a s u r e d Defect length (mm)

600

700

0.0

M e a s u r e d Defect length (nan)

Figure 7, Part B safe pressure, ASD format

For illustration, the safe working pressure according to the ASD format for the 45% deep measured defect is shown in figure 7, the full bold line labelled "Part B". An original design factor of 0.72 is assumed, resulting in a total usage factor of 0.9 x 0.72 = 0.65. Note that in this example the defect sizing uncertainty is not considered, and hence no additional factor included. The acceptance pressure using the calibrated safety factors is also included for comparison. As seen, if the defect sizing accuracy is reasonable accurate, the ASD format results in sufficient safety margin. When the inspection accuracy becomes poor, the safety margin becomes insufficient and an increased safety factor should be applied to account for the inspection uncertainty, which has to be judged by the user.

Figure 8, Acceptable measured defect size for various inspection accuracy at an operating pressure of 170 bar
Selecting the sizing accuracy of an inspection tool often includes a cost benefit evaluation, and the above information can support this evaluation, as the effect of the inspection accuracy is clearly shown. If defects of the order of 40-50 % depths of short pitting detects, or 30% for long defects, are expected, an accuracy of_+10% might not be sufficient for demonstrating the integrity of the pipeline at this pressure. The acceptable measured defects will increase for reduced operating pressure.

Alternative applications The form of the acceptance equation is made to determine the acceptable operating pressure for a measured corrosion defect in a pipeline. The equation can be re-arranged to determine the acceptable measured defect size for a specified operational pressure. The equation with the calibrated safety factors is used, and by omitting the factors the ASD equation can easily produce. By setting the specified operating pressure Poper equal to Pco,r, the equation can be re-arranged to:
(d/t)meas= 1 (l-Poper/P0) ed.StD[d/t ] (Eq. 8)

where P0 = Ym

2 t SMTS ( D - t)

(Eq. 9)

For a specific pipeline the measured relative defect depths that can be acceptable can be calculated for various operating pressures and inspection tool sizing capabilities. This can be very useful infurmation during operation, and also be a guide in selecting the required accuracy of an inspection tool. It is obvious that the more accurate the inspection tools can size the defects, larger measured

Single defect and longitudinal stress Compressive longitudinal stresses in a pipeline will reduce the burst capacity if the longitudinal stresses become significant. In DNV-RP-F 101 an interaction curve for the acceptable longitudinal stress and internal pressure for a pipeline with a corrosion defect is included. A few full-scale burst tests with combined loads were carried out, and the test results were supplemented by FE analyses (Bjsrnoy, 2000). During the development of the method it was discussed whether the burst test results followed a Von Mises like function or a Tresca like function, but the results did not clearly show any specific trend towards one of the functions. However, a Tresca like curve was selected for the allowable pressure for combined internal pressure and longitudinal compressive stress. The longitudinal stress in Figure 9 is the pipe wall stress. A benefit of the Tresca function is the horizontal plateau, where the calculated burst pressure is independent of the longitudinal stresses. The horizontal plateau was defined by P~orr, see equation 6, and the intersection with the x-axis remains to fully define the criterion. The interaction of the x-axis was defined as plastic collapse of the cross section of the pipe, and an appropriate safety factor was introduced. For an unconstrained (endcapped) pressurised pipeline the longitudinal stress is in tension, due to the end cap effect, and close to half the hoop stress. The interaction line can be defined by the equation, which describe the interaction criterion: 108

f Pcorr,comp = Pcorr ~l

cr Ltot A r "SMTS" ~ )

.~

(Eq. 10)

For boundary conditions other than end-capped, the basic criterion given in equation 10 should be used, alternatively update the equation in RP-FI01 in accordance with the applicable boundary conditions.
SUMMARY

The acceptable operating pressure is Pcorr,eomp according to the equation above, but limited to Por~.(Po~ as defined in equation 6) A~ is the remaining cross section at the corrosion defect, see RPFl01 for details.
internal pressure (or hoop stress)

,,,,~=r,, (o-,)(,_Z::(~/_R0~)

2, ,~a,~' O-r,,(a/,)*)

rr

mp P

aI,l~I

The background for the development of the DNV Recommended Practice, DNV RP-F101, for corroded pipeline has been outlined. The importance of the model uncertainty and inspection accuracy in the assessment of corrosion defects is demonstrated. The development of the interaction criterion is shown where a simple interaction equation becomes complex when excluding the contribution from the internal pressure to the longitudinal stress.
REFERENCES

A,.. 3)t4~'S.,~

Longitudinal stress O'Ltol

Figure 9, acceptance criterion for combined loading

However, the longitudinal stress in the x-axis for the approach above include also the resulting longitudinal stress due to the internal pressure, and since this pressure is the value to be determined, an iterative solution is required. Alternatively, the equation can be re-arranged by substitutes the total longitudinal stress in the Pcorr.compequation with, assuming endcapped conditions where the longitudinal stress is equal to half the hoop stress: D-t O't,tot = grL + Pcorr,comp(---~l

(Eq. 11)

The total longitudinal.stress O'Ltotin the pipe wall is the stress due to external loads OL and due to the end cap effect from the internal pressure Pcorr, comp(half the hoop stress). Inserting eq. I 1 into eq. 10 results in the following equation, I Pcon'.comp = Pcorr 1 D-t O'L + P corr,cornp ( - - - ~ 3 A r "SMTS(Eq. 12)

which can be solved for Pcorr, eomp as:

Pcrr'cmp =Pcrr

D4t -t

(Eq'13)

1 - Peorr Ar .SMTS. ~ and further re-arranged as the equation in DNV-RP-F101 : SMTS" ~ "~r(l _ Td (d / t) ,)

Pcorr,comp = Pcorr 1

rm

(Eq. 14)

2 " ~ ' A r (1. Yd(Q t ) * )

ASME, (1991), "Manual or Determining the remaining Strength of Corroded Pipelines, ANSI/ASME B31 Code for Pressure Piping". BaRe, A. D., Fu, B., Kirkwood, M. G., Vu, D. (1997), "New Methods for Determining the Remaining Strength of Corroded Pipelines", OMAE 1997. Benjamin A.C, Vieira R. D., Freire, J. L. and de Castro J.T.P, "Burst Tests on Pipeline with Long External Corrosion", IPC 2000, Calgary, Alberta, Canada. Bjornoy, O.H., Cramer, E.H., Sigurdsson, G. (1997), "Probabitistic Calibrated Design Equation for Burst Strength Assessment of Corroded Pipelines" ISOPE 1997, Honolulu. Bjornoy, O.H, Sigurdsson, G. Cramer, E.H., Fu, B Ritchie, D., (1999) "Introduction to DNV-RP-FI01 "Corroded Pipelines", OMAE 1999, Newfoundland, Canada. Bjornoy, O. H., Sigurdsson, G. and Cramer, E.H., "Residual Strength of Corroded Pipelines, DNV Test Results", ISOPE 2000 Chouchaoui, B. A. and Pick, R.J., (1992), "Residual Burst Strength of Pipe with internal Corrosion Pits", Proc. Int. Conf. On Pipeline Reliability, Calgary. Coulson, K. E. W, and Wolthingham, R.G. (1990), "Pipe Corrosion, Standard Damage Assessment Approach is Overly Conservative", Oil and Gas Journal. DNV Offshore Standard, DNV-OS-FI01, Submarine Pipeline Systems, Det Norske Veritas, 2000 DNV-RP-F 101, Corroded Pipelines, Det Norske Veritas, 1999. Fu ,B. and Kirkwood, M. G. (! 995), "Predicting Failure Pressure of Internally Corroded Linepipe using the Finite Element Method", Offshore Mechanics and Arctic Engineering, OMAE'95. Fu, B., Noble, J., (1998), "New failure pressure equation for corroded pipeline" Verbal presentation, OMAE 1998, Lisbon. Sigurdsson, E.H. Cramer, O.H. Bjorn~,, B. Fu, and D. Ritchie, "Background to DNV-RP-FI01 Corroded Pipelines", OMAE 1999 Jiao, G., Sotberg, T. and Igland, R. (1995), "SUPERB 2M STATISTICAL DATA: Basic Uncertainty Measures for reliability Analysis for Offshore Pipelines", SUPERB project report. Stephens, D. R. and Leis B., "Development of an Alternative Criterion for Residual Strength of Corrosion Defects in Moderateto High-Toughness Pipe", IPC 2000, Calgary, Alberta, Canada. BS 7910 "Guidance On Methods For Assessing The Acceptability Of Flaws In Fusion Welded Structures", BSI, London, UK, 1999.

109