Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Corrosion fatigue of steel catenary risers in sweet production (June 2008)

   

Richard Pargeter, David Baxter and Briony Holmes
TWI

Paper presented OMAE 2008 27th international Conference on Offshore Mechanics and Arctic Engineering, Estoril, Portugal, 15 - 20 June 2008. OMAE2008-57075.

Abstract

Steel catenary risers (SCRs) are commonly used for deepwater oil and gas developments and the most economic material of construction is generally carbon manganese (C-Mn) steel. These risers suffer cyclic loading principally due to vessel movements, and vortex induced vibration (VIV) from passage of marine currents. For this reason, close attention is paid to fatigue design and girth weld quality, and fatigue testing is commonly carried out on procedure test welds. A further advantage of C-Mn steel is that good quality welds can readily be made, and more importantly, freedom from defects can be assured by reliable ultrasonic inspection.

In sweet corrosive environments, when significant hydrogen effects would not be anticipated, a range of environmental effects on fatigue could be envisaged under different conditions, and at different stages of fatigue crack growth. For example, in early stages of growth, corrosion could blunt crack tips, and therefore slow the growth rate, whereas under other circumstances, or later in life, corrosion could provide additional crack extension, and accelerate growth.

It has been demonstrated in this programme of fatigue crack growth rate and endurance testing that the most aggressive conditions in terms of corrosivity may not give shortest fatigue lives in testing. The results of tests comparing behaviour in air and in a very highly corrosive aqueous environment at 60°C saturated with CO2 (conditions which could not be sustained in production) have been explained by reference to competing effects of fatigue and corrosion. Comparison has been made with other published data. Important safety implications surrounding conditions for project-specific corrosion fatigue testing for riser design are considered.

Introduction

If the hydrocarbon produced fluids are not corrosive, fatigue test data generated in air are appropriate for girth weld roots in steel catenary risers (SCRs)1. In such circumstances, standard fatigue design rules, and standard fatigue test techniques in air can be applied. The weld cap usually has to be dressed smooth to facilitate automatic ultrasonic testing (AUT), which also has benefits for fatigue strength, so this is of less concern than the root, even though it is operating in a seawater environment. Furthermore, there is a body of experience on the performance of C-Mn steel welds in seawater from offshore structures, and some work has also been done to explore the effects of this environment on SCRs.[1] The design of SCRs to handle corrosive produced fluids, however, has received much less attention. For corrosive conditions, where inhibition control is possible, C-Mn steel may still be the material of choice, providing corrosion fatigue behaviour can be determined and any issues arising can be resolved.

Work has been carried out at TWI and elsewhere which has explored the effects of H2S in the produced fluids on fatigue performance of welds in a range of materials (C-Mn steel, superduplex stainless steel, alloy 825 clad steel and titanium). In such sour environments, certainly at near-ambient temperature, corrosion fatigue behaviour of C-Mn steel will probably be dominated by hydrogen effects, rather than metal dissolution. However, in sweet corrosive environments, and/or at high temperatures in fluids with or without H2S, metal loss effects may be more important. Furthermore, different effects could be envisaged under different conditions, and at different stages of growth. For example, corrosion could blunt crack tips, and therefore slow the growth rate, or corrosion could provide additional crack extension, and accelerate growth. Furthermore, corrosion products or scale could limit fatigue crack closure. Any pitting or preferential weldment corrosion (PWC) would provide a stress concentration and encourage crack initiation, or conversely blunt crack-like features in this specific region. Thus, if the use of C-Mn steels is envisaged for sweet, corrosive service, a separate exploration of the effects of these environments on fatigue cracking is required. An initial assessment of these effects was the subject of a programme of work carried out at TWI for a group of Member companies.

It was evident from some of the findings that there was a risk of selection of unconservative corrosion fatigue test conditions for production testing, and in view of the safety implications, it was agreed that this part of the work should be published.

Materials and welding

A 254mm (10 inches) outside diameter 14.2mm wall thickness pipe complying with DNV OS-F-101 SML 450 IFU (approximately equivalent to API 5L X65) was procured and welded by Technip Offshore. The pipe was manufactured by Tamsa. Welding employed automatic GTAW root and GMAW hot pass fill and cap. A plain C-Mn Thyssen Union K52S wire was used for the root pass, and a low Ni Thyssen Nova Ni wire for the remaining runs.

A photomacrograph of weld BW1 is presented in Figure 1, with details of root toe profiles and HAZ, and weld metal microstructures in Figures 2 and 3 respectively. The root of specimen BW6-E1 after test is shown in Figure 4. The root width in these welds is such that a fatigue crack starting in this location would be expected to pass principally through HAZ microstructure. This should be borne in mind when comparing growth rate data (measured in weld metal) with endurance data, although relative frequency effects should be transferable. The HAZ has a soft, ferritic microstructure (Figure 2) and the weld metal consists principally of fine acicular ferrite (Figure 3).

Pargeter Figure 1
Fig. 1. Weld BW1 Etched in 2% nital. Magnifications given by mm scale
Pargeter Figure 2
Fig. 2. Details of root toes from weld BW1, showing profile and HAZ microstructure. Etched in 2% nital
Pargeter Figure 3
Fig. 3. Detail of weld metal microstructure from weld BW1 (approximately the centre of Figure 1a). Etched in 2% nital
Pargeter Figure 4
Fig. 4. Root of test specimen BW6-E1 after test

 

Environmental control and monitoring

General

A dedicated system for environmental monitoring and control during fatigue testing in this project was built. The deaerated test solution was continuously circulated between a reservoir (~22litre) and the test cell (~22litre) throughout test (Figure 5). A hot finger, incorporated in the recirculating loop to deposit iron (II) ions from solution, allowed control of the concentration of iron in solution, which would affect development of corrosion product films on the test sample.

Pargeter Figure 5
Fig. 5. Recirculation system for test solution

 

Continuous measurements of pH were made within the recirculating loop, and dissolved oxygen concentration was also measured at intervals during testing. Measurement of corrosion rate in the environment was conducted using Linear Polarisation Resistance (LPR) measurements on a separate specimen made from the pipe steel. Details of these measurements are given below.

The base line test environment was designed to give a high rate of corrosive attack (pH approximately 4.5). This consisted of 35000ppm NaCl and 68ppm NaHCO3 in water at 60°C, saturated with 100mol% CO2 at 1bara. This gives 13800ppm Na+, 21200ppm Cl and 50ppm HCO3 in solution. The calculated corrosion rate from models for this environment was 4.3mm/yr ±1.2mm/yr, and the scaling temperature was calculated to be 85°C.

pH measurements

The pH measurements were made using a commercial system consisting of an analogue pH probe connected to a digital gateway, which sent data to a controller, which was downloaded periodically. The probe was calibrated in pH4 and pH7 buffers prior to insertion in the test loop. Measurements were recorded every ten minutes. The data were corrected for temperature at the reference electrode automatically.

Iron measurements

Samples of solution were removed at intervals through the tests, and analysed for their dissolved iron concentration. This value therefore represents both the iron (II) ions in solution and the iron within corrosion product compounds that were held in suspension in the test fluid.

Oxygen and temperature measurements

Oxygen concentration in solution was measured directly using oxygen probes (Orbisphere brand) that were connected to the test vessels in their own flow loops. They were also used to record the temperature where this facility was available. Measurements were recorded every 30 minutes. The data were corrected for chlorinity and temperature based on guidance from the manufacturer.

Linear Polarisation Resistance (LPR) and corrosion rate data

A separate sample of parent material was inserted into a three-electrode set-up inside the test vessel for electrochemical monitoring. This working electrode was coated with two coats of Stopping-Off Lacquer No.45 (MacDermid Canning) - the same lacquer used on the test specimens - to leave ~1.6cm2 sample area (this was measured in each case). The reference electrode was a saturated calomel electrode (SCE), located outside the vessel, connected to the vessel electrolyte by a salt bridge. The auxiliary electrode was a piece of platinum wire.

A duplicate potentiodynamic sweep from -1500 to +100mV vs SCE at a rate of 1mV/s, was performed on the electrochemical samples in three of the tests following the end of each test. This was in order to generate data from which the actual Tafel constants for these systems could be measured. The specimen was left at the rest potential for 15 minutes between the sweeps. The measured and calculated Tafel constants were then used to generate the corrosion rates from the LPR data. The LPR data were gathered every 30 minutes.

Fatigue crack growth rate tests

Fatigue crack growth rate (FCGR) test equipment is shown in Figures 6 and 7. Tests were conducted using single edge notched bend (SENB) specimens in accordance with BS ISO 12108.[2] Specimens were extracted from the girth weld and notched at the weld root to allow growth in the through thickness direction, along the centre-line of the weld. Specimens had a square cross section (B=W=13.8mm), and a 1.3mm spark eroded notch to initiate cracking at the weld root. Standard servo-hydraulic testing machines with computer control and data logging were used, and crack length measurement was via the DCPD method.

Pargeter Figure 6
Fig. 6. Corrosion fatigue crack growth rate equipment (reservoir and recirculating loop not shown)
Pargeter Figure 7
Fig. 7. Corrosion fatigue crack growth rate specimen ready for testing

 

All tests were carried out under conditions of decreasing stress intensity range (ΔK), using an imposed stress intensity (K) gradient of 0.1mm-1, and a stress ratio (R) of 0.5. Tests were conducted at frequencies of 0.5Hz and 0.2Hz in the corrosive environment, and using a frequency of 10-20Hz in air. Four tests were performed in the environment and six in air. The purpose of the environmental tests was to assist with the choice of frequency for the endurance tests.

Fatigue endurance tests

Fatigue endurance tests were conducted under direct axial loading on strip specimens machined from the girth welds. Details of the test equipment and specimen design are shown in Figures 8-10. As a result of the curvature of the test section, axial loading produces a combination of membrane and bending stress at the test weld, with the weld root seeing the highest tensile stress range. Two strain gauges were placed each side of the weld root 10 and 25mm from the weld toe. Measured strains were extrapolated to provide an estimate of the local stress at the appropriate weld toe where failure was seen to initiate. The gauges were removed prior to commencing fatigue loading for the corrosion fatigue tests. The weld bead at the root was in the as-welded condition, while the cap was machined to preclude failure from this position.

Pargeter Figure 8
Fig. 8. Equipment for endurance tests in air
Pargeter Figure 9
Fig. 9. Equipment for corrosion fatigue endurance tests
Pargeter Figure 10a
Fig. 10. Specimen design for air
Pargeter Figure 10b
Fig. 10b. Specimen design for corrosion fatigue endurance tests

The endurance specimens tested in corrosive environments were also partially coated, principally to limit contamination of solution with corrosion products. The edges and cap surfaces were coated along the whole specimen length to preclude failure from these locations. The root side was coated to half way down the flared section, outside of the gauge length, as illustrated in Figure 11. This was done to minimise the risk of failures initiating at the edge of the coating.
Pargeter Figure 11
Fig. 11. Specimen BW6-E2 after test, showing extent of coating

 

Fatigue tests were conducted under load control with constant amplitude sinusoidal loading. The applied load was cycled down from a constant high maximum tensile stress (400MPa), to simulate the effect of residual stress in a full girth weld. Tests were continued to complete separation/maximum displacement, or to a target maximum endurance. Four specimens were tested in the environment covering an endurance range from around a hundred thousand cycles to about one million cycles.

Results

Environmental control and monitoring

The pH data are shown in Table 1. The pH at the start of the tests following addition of CO2 was pH4.0-4.7. During the test period, the pH generally rose by up to 0.8 pH units to between pH4.5 and pH5.4. In two cases (BW5-6 and BW3-E2) however, the pH dropped by 0.2 and 0.1 pH units respectively.

 

Table 1 In-situ pH data 

 Specimen Test type pH at start of test under test gas pH at end of test under test gas Duration of exposure to environment, days 
 BW5-5  FCGR  4.4  5.0  45
 BW5-8  FCGR  4.6  4.8  3.5
 BW5-6  FCGR  4.0  3.8  12
 BW5-9  FCGR  4.1  4.5  22
 BW6-E1  SN  4.5  5.1  56
 BW6-E2  SN  4.6  5.4  20
 BW3-E2  SN  4.7  4.6  2
 BW2-E2  SN  nr  5.4  9

 

The iron concentration data are summarised in Figure 12 and Table 2. As expected, the iron concentration in solution increased over the test duration, but in all cases scaling was noted on the hot finger (as expected), and inside the PVDF and PFA tubing used in the flow loop, suggesting that iron was being removed from solution through precipitation. In the case of BW6-E1, the concentration began to decrease between 28 and 40 days (672-960 hours) test duration.

Pargeter Figure 12
Fig. 12. Summary of results of analyses of iron in solution


Table 2a Results of analyses of iron concentration in solution (ppm) for FCGR tests

 Time/h BW5-5 BW5-8  BW5-6  BW5-9 
 0  0.3    0.7  
 29    3.2    
 30      0.4  
 72        1.1
 96      0.9  
 101    10.1    
 168        2.4
 264  8.2      
 282      2.1  
 408  15.6      5
 576  21.5      
 744  27.4      
 936  33.4      
 1080  33.7      

 

Table 2b Results of analyses of iron concentration in solution (ppm) for fatigue endurance tests

 Time/h BW6-E1  BW6-E2  BW3-E2  BW2-E2 
 0  2.2  <0.1     
 24      10  
 144  103.1      
 192        68.9
 264  170.7      
 288    79.7    
 360  199.7      
 432    97    
 435    95.1    
 672  244.4      
 960  157.4      
 1104  146.6      

 

Measurements using an Orbisphere gave high confidence that the oxygen content in solution was below 10ppb for the duration of all of the tests.

The temperature during the first test (BW5-5) was less stable than for subsequent tests (for which the control system was improved) and dropped from 60°C to 56°C by the end of the test, but was considered to have remained within an acceptable range. The temperature was successfully controlled to very close to 60°C for all other environmental tests.

Linear Polarisation Resistance (LPR), corrosion rate data and potential data

The calculated Tafel constants for this system were ba = 130mV and bc = 230mV. These were then used to determine the corrosion rates from the LPR data. The LPR and potential results are summarised in Table 3. The corrosion rate was ~6-7mm/year for all FCGR tests. The corrosion rates in the endurance tests in the environment were similar to those in the FCGR tests except for specimen BW6-E1. Specimens BW6-E1 and BW2-E2 had suffered lower corrosion rates (4-5 - 2-3mm/year) than that of BW6-E2 (6-7mm/year), and BW3-E2 (projected to be 6-7mm/year). This lower apparent corrosion rate was attributed to partial scaling of the LPR electrode.

The potential was generally very stable throughout all the tests (Table 3). It was approximately -715mVsce for all the FCGR specimens at the start of the tests. This rose to about -703mVsce on specimens BW5-5, BW5-6 and BW5-9, and to -675703mVsce on specimen BW5-8. For endurance specimens, the potential was between -735 and -690mVsce at the start of the tests. This rose by between 5 and 10mV on each of the specimens (BW6-E1, BW6-E2, BW3-E2 and BW2-E2).

 

Table 3 Corrosion rate and electrochemical potential results for FCGR and endurance tests.

 Specimen Test  Approximate corrosion rate at end of test, mm/yr (day this stable rate achieved)  Potential at start of exposure to environment, mV vs SCE  Potential at end of test mV vs SCE  Duration of exposure, days 
 BW5-5  FCGR  6 (13)  -715  -700  45
 BW5-8  FCGR  7 (2)  -710  -672  3.5
 BW5-6  FCGR  6 (0)  -715  -705  12
 BW5-9  FCGR  7 (0)  -715  -700  22
 BW6-E1  SN  3.5 (0) but 4-5 (12-32)  -690  -685  56
 BW6-E2  SN  6 (10)  -705  -695  20
 BW3-E2  SN  2 (1.5)  -735  -725  2
 BW2-E2  SN  3 (2)  -705  -707  9

 

Crack rate growth data

Baseline air data - The first test using specimen W01-01 was carried out under conditions of decreasing (ΔK) from an initial (ΔK) of approximately 300N/mm3/2. This test resulted in only a small amount of crack extension and relatively little test data were obtained. Subsequent specimens (W01-04 and W01-02) were then tested, again under conditions of decreasing (ΔK), this time at initial (ΔK) values of 400 and 600N/mm3/2 respectively. A summary of all of the test data is provided in Figure 13. It can be seen that test data are close to the mean curve for steels in air (R<0.5) taken from BS7910.[3]

Corrosion fatigue data - A summary of the corrosion fatigue crack growth rate data is included in Figure 13. The data obtained in air are plotted for comparison, and data are also compared to those reported by Szklarz[4] in Figure 14.

Pargeter Figure 13
Fig. 13. Summary of FCGR data
Pargeter Figure 14
Fig. 14. FCGR data compared with data from Szklarz

 

For the test carried out at 0.5 Hz (BW5-6), the initial ΔK was 800N/mm3/2 and the test lasted about eight days (from start of cycling). Initially (ΔK=800) the growth rate was somewhat higher than might be expected in air. However, as the applied ΔK decreased (K= 500-600) the crack growth rate was comparable to or lower than that seen in air.

Specimen BW5-9 was tested using a lower frequency of 0.2 Hz, and the initial ΔK was again 800N/mm3/2. At high ΔK the crack growth rate was similar to the previous test (just under 2x10-4 mm/cycle), but at lower ΔK, the crack growth rate was noticeably higher than for BW5-6. This test lasted approximately three days.

Endurance data

Baseline air data - Tests were performed in air at a frequency of 10Hz. Three specimens were tested to failure at local stress ranges of 174, 206 and 303MPa. Three further specimens tested at local stress ranges of 99, 161 and 167MPa, ran out at >5x106 cycles. A summary of the test data is given in Table 4 and included in Figure 15. It can be seen that all of the data are to the right of the mean Class E curve taken from BS7608.[5]

Pargeter Figure 15
Fig. 15. Summary of fatigue endurance test data


Corrosion fatigue data - Four endurance tests were conducted in the corrosive environment. Test parameters and results are summarised in Table 4 and Figure 15. The test frequency for the first two of these tests was 0.2Hz.

 

Table 4. Endurance test data

 Specimen Environment Frequncy, Hz  Local stress range, MPa  Endurance cycles  Failure location 
 BW1-A1  Air  6  206  477,265  Weld root toe
 BW1-A2  Air  6  303  184,435  Weld root toe
 BW3-A1  Air  6  174  778,555  Weld root toe
 BW3-A4  Air  6  161  4,820,427  Run-out
 BW2-A4  Air  6  167  6,098,795  Run-out
 BW2-A3  Air  6  99  10,300,000  Run-out
 BW6-E1  Solution  0.2  301  942,081  Parent metal (within grip and guage length)
 BW6-E2  Solution  0.2  380  312,724  Run-out
 BW3-E2  Solution  2.5  300  90,378  Weld root toe
 BW2-E2  Solution  2.5  200  1,720,670  Parent metal (within grip)

 

Specimen BW6-E1 was tested at a stress range of 300MPa and was on test for 942,081 cycles (55 days), that is, longer than the fatigue life in air at this stress range. Upon removing the specimen from the test cell, it was clear that the displacement trip had in fact been triggered by development and extension of a crack in one of the pin loaded ends (outside the test cell), but there was also a small crack within the specimen gauge length. This smaller crack was remote from the weld (~20mm from the weld toe), as illustrated in Figure 16 and 17.

Pargeter Figure 16a
Fig. 16a. Specimen BW6-E1 after test (942,081 cycles, 57 days exposure: a) Full specimen
Pargeter Figure 16b
Fig. 16b. Specimen BW6-E1 after test (942,081 cycles, 57 days exposure: b) Central region, showing weld root, and crack (arrowed)
Pargeter Figure 16c
Fig. 16c. Specimen BW6-E1 after test (942,081 cycles, 57 days exposure: c) Surface of crack after breaking open
Pargeter Figure 17
Fig. 17. Section through specimen BW6-E1 after test (942,081 cycles, 57 days exposure). Etched in 2% nital a) Weld toe (in the as welded condition prior to test) b) Primary crack position (dressed smooth prior to test)

The second test specimen (BW6-E2) was tested at a stress range of 380MPa and was on test for 312,724 cycles (18 days). At this point, the fatigue life had again exceeded that expected in air by a considerable margin, so the test was terminated prior to failure. The appearance of the test specimen after testing is illustrated in Figure 11, with a detail of the central region in Figure 18.
Pargeter Figure 18
Fig. 18. Detail of central region of specimen BW6-E2 after test. See also Figure 11

 

Data from these two tests suggested that lives in the sweet corrosive environment were significantly higher than those seen in air, and were approximately a factor of eight higher than the mean for Class E. It was decided that a complete series of six tests in this environment was not necessary.

Effect of frequency

Further testing in this environment was carried out using a cyclic frequency of 2.5Hz to investigate the influence of this test parameter. Specimen BW3-E2 was tested at a stress range of 300MPa to provide a direct comparison with specimen BW6-E1 described above. Failure of this sample occurred after 90,378 cycles (1 day), an order of magnitude lower than the 0.2Hz test, and a factor of 2 lower than the corresponding test carried out in air. Examination of the specimen after the test revealed that failure had occurred from the weld root toe.

The effect of frequency at a lower stress range was explored using specimen BW2-E2, at a stress range of 200MPa and a frequency of 2.5Hz. This specimen was stopped at 1,720,670 cycles, beyond the air results at this stress range, due to cracking in the pin loaded end. Thus, at this stress range, increasing the frequency did not change the beneficial environmental effect on life observed at 0.2Hz. This is discussed below.

Discussion

General comments

Prior to the start of this project, it was recognised that there could be a number of competing factors affecting both fatigue endurance and fatigue crack growth rate in sweet corrosive production environments [Woollin et al]. At ambient temperature, and in sour environments hydrogen effects generally dominate, and relatively simple environmental monitoring (pH, H2S content, temperature) have been sufficient to generate confidence in the conditions of exposure and experimental control. In the present work in sweet conditions, where testing was at elevated temperatures, more detailed and extensive monitoring have been required. The purpose was to gain a more detailed understanding of the effect of the environmental variables on fatigue crack initiation and growth, which is believed to be primarily due to direct corrosion effects under these conditions, rather than hydrogen effects. The corrosion mechanism is primarily anodic dissolution of material in sweet conditions and is significantly influenced by development of corrosion product films and complex interaction of compounds in the environment (solubility of iron, formation of protective or semi protective scales etc). Therefore, monitoring of pH, iron concentration in solution, oxygen concentration, temperature, electrochemical potential, and the measurement of corrosion rate (LPR in a conductive solution) were carried out.

The in-situ oxygen sensors, pH electrodes and LPR probes have proved successful in providing assurance of the environmental conditions during testing, and the corrosion rate data have proved particularly useful in developing an understanding of the endurance results as detailed below. The measurements of iron in solution have provided an indication of the likelihood of scaling in some of the longer endurance tests.

Fatigue crack growth rate

The original purpose of the fatigue crack growth rate testing was to determine an appropriate test frequency for the endurance testing. In the event, this was not successful, as it is now believed that the processes involved in fatigue crack initiation and growth, at least growth in deep cracks, respond differently to the presence of a corrosive environment, and that deep crack behaviour does not dominate endurance. Nevertheless, useful data were generated.

With regard to frequency, there was a slight effect with 0.2Hz (the lowest frequency used) providing the highest growth rate. The spread in data was not large, however, and in summary, the growth rate was higher than in air at ΔK over about 600N/mm3/2 (Figure 13) and it also appeared to be lower than in air at lower ΔK.

As the tests were carried out under decreasing ΔK conditions, it is possible that a premature threshold had been reached due to the onset of crack closure associated with the build-up of corrosion products within the crack. Similar data generated by Szklarz[4] also show this effect. In that work, tests were carried out using 'blocks' of constant ΔK, each block being at 5-10% lower ΔK than the previous one, so that the possibility of crack closure effects cannot be entirely ruled out for these data either. Nevertheless, despite a number of experimental differences, there is very good correspondence between the two data sets at 0.2Hz (Figure 14) and both sets of data show a frequency effect which is more marked at lower ΔK. This suggests that there is a real effect.

The mechanism for increasing growth rate in such environments, which has been proposed to date is crack tip dissolution. Szklarz found a marked increase in growth rate at lower ΔK (~400-650N/mm3/2) when the frequency was reduced to 0.04Hz, and under these circumstances the growth rate was higher than air, with a similar slope to the air data, (provided cracking did not follow the fusion boundary, in which case it was considerably higher) (Figure 19). Because of the frequency effect, Szklarz attributed the higher growth rate to crack tip dissolution. Under such circumstances, a constant difference in da/dN would be anticipated between air and environment, regardless of ΔK, such that:

Pargeter Formula 1

and this is what is observed for Szklarz's data up to 635N/mm3/2. (Two bounding lines, air + maximum observed difference in growth rate = 8.01 e-5 mm/cycle and air + minimum observed difference in growth rate = 2.73 e-5 mm/cycle have been superimposed in Figure 19).
Pargeter Figure 19
Fig. 19. Evaluation of the effect of a constant rate of crack tip dissolution on fatigue crack growth rate for data from Szklarz

 

Having said that, if the crack tip dissolution effect determined in this way for 0.04Hz is factored and applied to the 0.2Hz data (giving maximum and minimum additional growth rate components of 1.60 e-5 and 5.45 e-6 mm/cycle respectively), some acceleration would also be expected, as shown on Figure 19, but was not observed from Szklarz's paper. (No measurable acceleration would be expected on this basis at 1Hz.)

At ~800-1000N/mm3/2, this relationship no longer holds (although there are no air data at this stress intensity range), and Szklarz proposes some unknown accelerating mechanism other than crack tip dissolution. The correspondence with TWI data suggests that this is a real effect, and that this is possibly similar to stress corrosion fatigue[6], rather than pure corrosion fatigue, but there is no ready explanation for the mechanism. This is not of too much concern, however, as growth at such high ΔK would only be experienced close to end of life.

Szklarz suggested that low crack growth rates in his uninhibited brine at low ΔK were due to scale build up and crack closure. It is of interest to note that the conditions in Szklarz's work would be expected to generate a carbonate scale, whereas the present environment was selected to be highly corrosive, but non scaling. Thus, it is not clear that this explanation can be used for the present TWI data, even though the performance is closely similar.

Endurance testing

The results of endurance tests were initially surprising, as the lives considerably exceeded the performance in air, to such an extent that failures were experienced in the specimen ends rather than at the weld. It is believed that the explanation lies in the competition between corrosion rate and fatigue crack growth rate; if material is being removed from the surface at a rate faster than the initial fatigue crack growth rate, then development of a fatigue crack will effectively be prevented, provided the corrosion mechanism does not lead to the formation of sharp features. This is similar in concept to the model proposed by Chen et al[7] who considered competition between fatigue cracking and pitting. In this model, a dual criterion for cracking is proposed, namely

Pargeter Formula 2
And
Pargeter Formula 3

Which is equivalent to

Pargeter Formular 4

 

In the present case of fatigue of a weld, the relevant competition is between fatigue cracking from an inherent weld toe intrusion and corrosion in that region. In the corrosive environment, the observed corrosion rate is approximately 7mm/year, so:

  • At a frequency of 0.2Hz this equates to approximately 1nm/cycle.

  • At a frequency of 2.5Hz this equates to approximately 0.1nm/cycle.

 

In order to model the behaviour of an endurance test specimen some assumptions need to be made regarding the initial flaw size. For the current analysis two potential geometries for a typical weld toe intrusion are considered:

  • 0.13mm x 0.43mm (case 1).

  • 0.07mm x 0.7mm (case 2).

 

For an applied stress range of 200MPa this results in an applied ΔK of 184N/mm3/2 (case 1) or 171N/mm3/2 (case 2). Similarly for an applied stress range of 300MPa this results in an applied ΔK of 276N/mm3/2 (case 1) or 257N/mm3/2 (case 2).

Anticipated crack growth rates can then be calculated from a given da/dN : K relationship. In this instance the BS7910 mean curve for steels in air (R<0.5) has been chosen, as experimental data were seen to lie close to this BS7910 curve over the range of K where measurements were made in air. The table below summarises the analysis.

 

 Stress range Defect size  ΔK  da/dN (crack)
 200MPa  0.13mm x 0.43mm  184  0.03nm/cycle
   0.07mm x 0.7mm  171  0.02nm/cycle
 300MPa  0.13mm x 0.43mm  276  1nm/cycle
   0.07mm x 0.7mm  257  0.5nm/cycle

 

It can be seen that at a stress range of 300MPa the predicted crack growth rate (0.5 - 1nm/cycle) is higher than the corrosion rate at 2.5Hz. Therefore fatigue (or corrosion fatigue) cracking would be expected, as was indeed the case. However, the crack growth rate is less than (or equal to) the corrosion rate at 0.2Hz, so corrosion would be expected to dominate and no cracking to occur, again, as observed.

At a stress range of 200MPa the predicted crack growth rate (0.02-0.03nm/cycle) is less than the corrosion rate, even at 2.5Hz, so corrosion would be expected to dominate in both instances, as was indeed the case.

At 2.5Hz and 300MPa stress range the environment has apparently had a slight adverse effect on fatigue life. Although more data would be required to confirm this, the measured endurance at 2.5Hz was 64,158 cycles, which is approximately a factor of 2 lower than seen in air, and is similar to that reported by South West Research Institute as indicated in the earlier literature review.[8] Thus corrosion is apparently providing some slight acceleration of growth rate under these conditions.

A consideration that has not been made, in the above assessment, is the effect of corrosion pitting on fatigue crack initiation. It was evident that pitting did develop over time, and that this could develop into fatigue cracking. Sample BW6-E1, which was exposed for 57 days and survived 942,081 cycles, failed at the pin loaded end outside the environment, but was close to failure from a fatigue crack which had developed from pitting in an originally polished surface (prepared for strain gauging) ~20mm from the weld toe, (see Figure 17b).

Thus, the true situation is probably not a simple competition between uniform corrosion and fatigue propagation, but also involves pitting. In the early stages of pitting, pit growth will be faster than fatigue propagation from the slight stress concentration due to the pitting. As the pitting develops, the rate of pit growth typically slows down, but the severity of the stress concentration, and thus fatigue growth rate, is higher. This is in essence Chen et al's model, referred to above, but will be more complicated in the presence of a pre-existing weld toe stress concentration and in situations where corrosion rate is influenced by local tensile stress differences (stress assisted corrosion).

Application of results

As originally envisaged at the outset of this work, different effects under different corrosive conditions, and at various stages of growth could increase or reduce fatigue crack growth rate Thus, the results of this project were not entirely unexpected. Consideration of the interactions between fatigue and corrosion processes above has allowed an understanding of the observed effects to be developed and has indicated the complexity faced in extrapolating to other conditions.

Clearly the environment and loading conditions explored and the number of data points generated is insufficient to allow data to be applied directly to any project situation, and indeed such highly corrosive conditions will not occur, at least for any significant duration, in practice. The success of this present project, however, is that it has demonstrated how easily significant errors in the identification of worst case conditions can be made. Fatigue testing under worst case corrosive conditions could provide unconservative design data. Until there is a much improved understanding of all the influencing factors, safe design fatigue data must be generated in project specific testing, using conditions which are close to the intended service. Further research to broaden the basic understanding to a wider range of conditions, such as loading frequency, corrosive environments, solution conductivity, inhibitors and aspects such as fluid flow impact on product film and inhibition (transport and residence) is required to reduce this need. Although the knowledge gained in this work will contribute to this understanding, and reduce the risk to some degree, its main benefit in isolation is to demonstrate the potential pitfalls.

Having decided on test conditions, decisions also need to be made on whether any pre conditioning is appropriate. If there is a risk of pitting, the early stages of a test may be taken up with pitting, which is principally dependent on time rather than number of cycles, and an endurance (number of cycles) which does not take account of this initial period may be unconservative. Where pitting may be a factor, close attention will also need to be paid to correctly modelling pipe surface condition in testing.

Other practical considerations for development projects include the selection of inhibitors in consideration of fatigue, and for operating assets there is a need to be certain that changing inhibitor (or simply active constituents) during the operational life is supported by data to confirm efficacy in terms of a pre-corroded system and fatigue.

Conclusions

This work has demonstrated the difficulty of identifying worst case conditions for corrosion fatigue testing, and the importance of a detailed knowledge of the environmental conditions during test. For project specific testing, it is therefore important to model and monitor the principal environmental variables as closely as possible.

For the highly corrosive conditions used in this testing, the following specific conclusions have been drawn. However, care needs to be taken not to apply these trends to operating steel catenary risers as such highly corrosive conditions would not be acceptable for anything but the shortest periods in service and the trends are likely to be different for less corrosive conditions:

  1. Fatigue crack growth rate was higher than in air at higher ΔK (over about 600N/mm3/2) but slower at lower ΔK.

  2. There was only limited effect of frequency on fatigue crack growth rate, but it was apparent that highest growth rate (mm/cycle) was at the lowest frequency.

  3. Fatigue endurance data for air tests lay just to the right of the Class E curve taken from BS7608, and showed an apparent fatigue limit at around 160MPa stress range.

  4. Fatigue endurance at 0.2Hz exceeded air data by a factor of about 8.

  5. Increasing the frequency by an order of magnitude changed the environmental effect at 300MPa, such that the endurance became a factor of 2 lower than the air results. No effect of the same increase in frequency was discernable at 200MPa stress range.

  6. Increases in fatigue endurance are believed to be due to a competition between material removal by general corrosion and fatigue crack growth rate, such that fatigue crack growth is hindered under highly corrosive conditions.

  7. Where reductions in fatigue endurance were observed, they are believed to have been due to crack tip dissolution.

 

Acknowledgments

The work has been supported by the following companies and both their support and permission to publish are acknowledged with gratitude:

  • BP Exploration Operating Company

  • Chevron Energy Technology Company

  • PETROBRAS

  • TOTAL S.A.

  • Tubos de Acero de México S.A. (Tamsa)

  • Technip Offshore UK Ltd.

  • Shell International Exploration & Production Inc.

 

References

  1. Razmjoo GR, Hadley I and Crouch S. 'Fatigue of TLP (tension leg platform) tendon girth welds'. Proceedings, 6th International Offshore and Polar Engineering Conference, Los Angeles, USA, 26-31 May 1996. Vol.4. pp.190-198.

  2. BS ISO 12108:2002 'Metallic Materials - Fatigue Testing - Fatigue Crack Growth Method'. British Standards Institution.

  3. BS 7910:2005 'Guide to methods for assessing the acceptability of flaws in metallic structures'. British Standards Institution.

  4. Szklarz, KE: 'Aggressive CO2 corrosion and fatigue behaviour of pipeline girth welds'. Corrosion 2002, paper 00012 NACE.

  5. BS 7608:1993 'Code of practice for fatigue design and assessment of steel structures'. British Standards Institution.

  6. Woollin P, Pargeter RJ and Maddox SJ: 'Corrosion fatigue performance of welded risers for deepwater applications'. Corrosion 2004, paper 04144 NACE.

  7. Chen GS, Wan K-C, Gao M, Wei RP and Flournoy TH: 'Transition from pitting to fatigue crack growth - modelling of corrosion fatigue nucleation in a 2024-T3 aluminium alloy'. Materials Science & Engineering, A219 (1996), pp.126-132.

  8. Lee BK: 'Corrosion fatigue of steel Catenary risers in sweet production - literature review' TWI report 15474/1/05 May 2005.

For more information please email:


contactus@twi.co.uk