HYDROTESTING EFFECTS-CONCLUSION TESTS SUPPORT EFFECT OF LINE PIPE TESTING ON CRACK GROWTH
B. N. Leis, F. W. Brust
Battelle Columbus Division
Columbus, Ohio
Full-scale tests of line pipe with patched through-wall flaws and part-through-wall flaws indicate that significant crack growth is possible as a result of repeated pipeline hydrotesting.
These test results show that such growth may involve a significant fraction of time-dependent cracking and can cause failure at pressures well below those predicted by current methods.
Predictions based on the engineering model presented in Part 1 of this series (OGJ, Feb. 12, p. 45) reasonably matched observed flaw-growth behavior and failure pressure.
The model, based on J-tearing theory, was developed to assess how flaw growth during hydrotesting influences line-pipe integrity.
EXPERIMENTAL ASPECTS
The approach taken was to develop an engineering model of time independent and time-dependent cracking stepped from simple laboratory flaw and specimen geometries through to line-pipe applications.
Validation at the laboratory scale, which made use of test specimens with different loading or flaw sizes than were used to characterize material properties, has been reported .4 (References appear at the end of Part 1.)
Studies were then made on full-scale pipe sections designed to evaluate the model's ability to predict flaw growth and limit pressure.
The material involved in both full-scale tests was nominally X52, rolled in the early 1970's.
The first test was on a section of 22-in. OD X 0.270-in. W.T. pipe designated as DN3. The second and third tests were on sections of 36-in. OD X 0.394-in. W.T. pipe designated as CR9.
Values of maximum operating pressures (MOP) and the pressures corresponding to Pt/P, of 1.0, 1.25, and 1.5 are 930 psi, 1,165 psi, and 1,398 psi for DN3. These same values for CR9 are 828 psi, 1,036 psi, and 1,243 psi, respectively.
The chemistry of both pipes met the API 5L X52 specification. The 0.5% total strain yield stress of both steels was between 59 and 60 ksi. Both steels had ultimates in excess of the API minimum of 72 ksi. The elongation of CR9 was slightly greater than that for DN3, despite the fact that full-thickness (estimated) CVN (Charpy V-notch) energies for transverse specimens were 30 ft-lb for CR9 and 32 ft-lb for DN3.
All data developed were representative of X52 line pipe. Therefore, it is expected that flaw growth in these pipe sections will be representative of X52 pipe in service.
Test pipes were prepared in an identical manner. Flaws were located centrally in endcapped pups with the flaws separated from the caps by about 2D. For DN3, three through-wall (TW) flaws were cut along a circumference about 120 apart, whereas for CR9-1 four TW flaws were cut on a circumference about 90 apart.
TW flaws were sharpened with a jeweler's saw for the last 0.25 in. For DN3, the flaw lengths were 0.75, 1.50, and 2.50 in., while for CR9-1 the lengths were 1.0, 2.15, 2.50, and 3.1 in.
Part-through-wall (PTW) flaws were cut into the outside diameter of CR9-2 locations similar to CR9-1. These flaws had lengths of 4.32, 4.74, 4.98, and 5.27-in. with aspect ratios of 0.052, 0.056, 0.061, and 0.07, respectively.
Flaw sizes were estimated to exhibit no growth, some growth, and significant growth to permit the broadest evaluation possible. After the TW flaws were cut, composite patches consisting of layers of stainless steel backed by soft aluminum and rubber were glued to the inside diameter, well removed from the flaw; the ends were capped and nipples installed.
The pups were instrumented with strain gauges, crack-mouth-opening displacement (CMOD) gauges, and electric potential (EP) probes to measure flaw growth, and signal conditioning was completed.
The loading consisted of a sequence of pressure steps each of which was followed by a hold time. The procedure involved an initial pressure step of several hundred psi followed by smaller steps in which the hold time was set based on the time for any creep apparent in strain gauges or clip gauges or in EP signal to decay. This process continued until a leak precluded further pressurization.
RESULTS, MODEL EVALUATION
Typical results from the instrumentation are shown in Fig. 1, each part of which shows strain gauge, CMOD gauge, and EP and crack length vs. pressure, respectively. These data represent the largest flaw in DN3.
The results in Fig. la show longitudinal and circumferential strain vs. pressure at a location two flaw lengths from the end of the flaw on a line 1 in. off the longitudinal axis of the flaw. Geometric nonlinearity due to the flaw opening and related bulging are apparent in the longitudinal strain and in the circumferential strain up to about 1,250 me,
Beyond 1,250 me, the circumferential strain also exhibits material nonlinearity due to creep and plasticity. Portions of the strain history that show strain increasing without a pressure increase represent creep-related deformation and time-dependent extension. Note that beyond about 1,000 psi, the greatest changes in strain are time dependent, indicating that time-dependent deformation dominates time-dependent (plastic) contributions.
This tendency increases strongly as the pressure approaches the leaking condition, which for the flaw size selected develops before the 90% SMYS state can be achieved. For this flaw case, albeit somewhat unrealistic, creep dominates the observed strain-gauge response between maximum allowable operating pressure (MAOP) and the attempt to reach the hydrotest level of 90% SMYS.
Fig. 1 b presents CMOD vs. EP and pipe-pressure data. The slope change in EP vs. CMOD is indicative of crack initiation which for the example shown occurs at about 850 psig. The pressure-vs.CMOD trend shows steps similar to those observed in the strain-gauge records indicating that beyond about 900 psig most of the CMOD occurs by a time-dependent (creep) mechanism.
Beyond about 1,050 psig, the increment in EP (AEP) which is proportional to crack extension, becomes dominated by a time-dependent mechanism. Crack length vs. pressure derived from EP records, such as presented in Fig. 1 c, clearly shows this tendency.
Beyond initiation and up to MAOP, the growth is pressure dependent and largely time independent (i.e., plasticity controlled). Beyond MAOP and up through leaking, the crack-growth process becomes increasingly time dependent.
The results presented in Fig. 1 point to a crack-growth process that under hydrotest conditions is strongly time dependent for through-wall flaws. It is significant that the stiffness of this pipe was such that a TW flaw 2.5 in. long did not grow delectably at pressures up to 90% MOP and that, once active, the flaw growth was slow and stable.
Analysis of limited data from the PTW full-scale test (CR9-2) has indicated that the constraint developed at the tip of the PTW flaw reduces the magnitude of the time-dependent extension to levels comparable to those of the time-independent effects up through 90% SMYS.
ACTUAL, IDEAL HISTORIES
Fig. 2 summarizes the fullscale data on flaw growth vs. pressure for the largest flaw in DN3, shown as dotted trends along with various predictions of the flaw-growth behavior. The black dot located along the pressure axis is the failure-pressure prediction (1,313 psi) based on the currently available method of Maxey.14
The corresponding initiation pressure is 633 psi. Initiation is predicted with the present model at 680 psi. If only a time-independent analysis is performed, failure is predicted at 1,273 psi after about 0.28-in. extension at either flaw tip.
Results of two time-dependent analyses are also presented. One is shown as a solid line marked "actual history," and the second is shown as a dashed line marked "idealized history."
The predictions for actual history represent analysis that tracked the exact pressure-time history in the test. The second analysis tracked an idealized pressure-time trend that involved fewer pressure steps and longer hold times that would reduce analysis time and cost.
Both time-dependent predictions led to a failure pressure of 1,100 psi, with crack extension of about 0.34 in.
Comparison of the predictions with the actual data indicates a very close match between the time-dependent analysis (which also includes the effects of time-independent cracking) and the actual behavior for both crack growth and failure pressure.
For this case, there is little difference between the idealized history and the actual history indicating some time and cost savings may be gained in this manner. There is, however, a significant difference between test and predicted results when time-dependent effects are ignored-a discrepancy that apparently will be less when PTW flaws are dealt with.
Not surprisingly, the worst prediction occurs when all sources of flaw extension are ignored. Failure pressure is nonconservative by about 16% in this case and corresponding flaw sizes estimated via LEFM would be smaller than actual sizes by about 0.5 in. at the point of maximum pressure.
Such flaw growth during hydrotesting reduces the anticipated margin of safety based on flaw size and can cause pressure reversals.
Consider now the comparison of predicted and observed trends for CR9 shown in Fig. 3. This test involved an initial load through seven pressure-hold steps to a maximum pressure of 1,050 psi at which time the test was aborted.
TWO-STAGE ANALYSIS
Subsequently, after repatching, the test was restarted and continued through pressurize-hold steps to failure. Thus, this test has been analyzed in two stages.
The first considered the possibility of initiation during the aborted segment. This analysis predicts initiation at 600 psi whereas the experiment showed evidence of initiation at 760 psi. The flaw was predicted to grow about 0.62 in. during this time.
The second analysis stage, results for which are shown in Fig. 3, ignored the initial history and sought to analyze flaw extension of the initial flaw because clear evidence of flaw growth could not be observed at the outside diameter of the pipe prior to restarting of this test.
Results for predicted flaw growth for the PTW flaws in test CR9-2 are shown for example in Fig. 4 for the two shortest flaws. Trends are similar to those for the TW flaws.
In both cases, the depth predicted at leaking which occurred at 1,050 psi was within several percent of the observed growth.
The comparison of predictions and observations (Figs. 3 and 4) indicates trends similar to that for DN3. Predictions either ignoring crack growth or not including time-dependent effects significantly overestimate the failure pressure and underestimate actual flaw extension.
The failure-pressure prediction which uses the current model slightly underestimates the actual pressure for TW flaws, whereas results for flaw growth match reasonably but underestimate observed PTW results.
Significant flaw growth is predicted for both PTW and TW flaws but what was actually observed is always underestimated.
Work has been under way to modify the prediction procedure to account for history effects, such as crack growth in prior hydrotests, that have heretofore been ignored.
HYDROTEST HISTORY
The formulation of this model in conjunction with the materials data developed implies that flaw growth in ductile line-pipe steels depends on the time and cycle-dependent nature of the hydrotest history.
A related trend is that certain hydrotest histories could reduce the failure pressure of the line if these histories were repeated. Likewise, stress reversals could develop depending on the conditions used in repeated hydrotesting.
While none of these implications can as yet be proved, they are now somewhat better substantiated in that the model has been shown to produce viable predictions for full-scale tests of TW and PTW flaws. Further validation of the model will come through its application to another series of full-scale test.3 Such studies are currently under way.
In a more practical light, the data presented, when coupled with the model, suggest that the nature of the loadtime history used in the hydrotest may either promote or retard cracking, depending on defect sizes and pressure levels. Establishing the hydrotest history, which is both practical and causes the least possible crack extension, remains as perhaps the most important use of this model.
MODELING FLAW GROWTH
The results of this study lead to a number of important conclusions concerning flaw growth in line pipe under hydrostatic conditions. These conclusions relate to the ability to model flaw growth and the implications of the time and cycle-dependent flow properties of the material.
Conclusions regarding crack-driving force include:
- Time-dependent flow, such as observed in the material studied, can cause an increase in the crack-driving force (J).
- An increase in crackdriving force (J) can be expected as a consequence of cyclic load and hold-time effects, such as are involved in a hydrotest.
- Mechanical prestrain tends to reduce the cracking resistance.
Conclusions concerning the behavior of line pipe subjected to overload sequences, such as a hydrotest, include:
- Good prediction of ductile cracking behavior in fullscale tests of flaws in X52 line pipe can be achieved with an uncomplicated analysis procedure.
- Certain combinations of lead and time can be determined which facilitate practical hydrotesting with the least potential for crack growth.
- Repeated hydrotesting can cause cracks to grow at pressures less than achieved in a prior test.
Copyright 1990 Oil & Gas Journal. All Rights Reserved.