Sustained load cracking of titanium alloy weldments (June 2005)

Authors are with TWI Ltd, Granta Park, Cambridge CB1 6AL, UK.

Proceedings of OMAE 2005: 24th International Conference on Offshore Mechanics and Arctic Engineering (OMAE 2005), June 12-16, 2005, Halkidiki, Greece

Abstract

Failure of critical titanium parts, including some offshore components, has drawn attention to delayed cracking in Ti-6Al-4V alloys, but, given good design and alloy variant selection, such failures are avoidable. Delayed cracking, or sustained load cracking (SLC), can occur at low to moderate temperature (approximately: -50 to 200°C), depending on the titanium alloy and condition. Appropriate testing methods are required to generate stress intensity threshold values (K _ISLC ) that can be incorporated into the design of titanium structures and recommendations are needed on the optimum chemistry and microstructure for greatest resistance. In the present work threshold stress intensity factor data (K _ISLC ) were generated for Ti-6Al-4V alloy sheet, forgings, pipe and weldments using two different rising stress intensity factor test methods. It is concluded that material with a beta-annealed microstructure and low oxygen content (i.e. extra-low interstitial material such as ASTM Grades 23 and 29), has high resistance to SLC and that weld metal and transformed heat-affected zone also perform well, before and after postweld heat treatment, provided interstitial element pick-up during welding is prevented. Purchasing material in a general 'mill annealed' condition is not recommended without specifying acceptable microstructures. Further refinement of test method is also recommended for defining K _ISLC .

Introduction

Titanium alloys are susceptible to time-dependent failure (sustained load cracking, SLC) if stresses or stress intensity factors exceed certain threshold values. Failures in critical titanium alloy components such as aero engine fan blades and certain offshore components have drawn attention to sustained load phenomena in these materials, be they 'cold creep', sustained load cracking, dwell fatigue or ripple fatigue. Experience indicates that, for Ti-6Al-4V alloys, susceptibility to the most damaging sustained load phenomena is likely to be greatest for the following situations [1]:

• Temperatures between 0 and 25°C.
• Microstructures typical of mill-annealed products.
• Materials with high aluminium and/or oxygen contents.
• Products with some degree of crystallographic texture.
• Materials with high hydrogen content (typically >150ppm), i.e. outside most material specifications.

Data are scarce and are most needed to facilitate design for the avoidance of SLC in structures made from the most common alloy, Ti-6Al-4V. Any study must consider the variability inherent in Ti-6Al-4V that can result from differences in chemistry, heat treatment and processing route. Weldments are also clearly a concern, due to the potential for greater stress intensities at weld flaws and the stress concentrations presented by the weld toes. Thus, for the present work, material and weldments from selected forgings, plate and pipe were investigated. Standard (ASTM Grade 5) and extra-low interstitial (ELI, ASTM Grade 23) materials were examined with examples of typical mill-annealed and beta-annealed batches being included.

One of the difficulties in determining the threshold stress intensity factor, below which SLC will not occur, K _ISLC , is in the choice of test method. Two different strategies of testing are available, based on either rising or falling K. In a rising K test, a fracture mechanics specimen is held under a constant load. If the crack grows, then the K increases, resulting in accelerated crack growth. This is representative of components that experience a given load as opposed to a fixed displacement. Falling K tests are based on fracture mechanics specimens held under a fixed displacement. This type of test suffers from several problems, including that relaxation as a consequence of sustained load strain [1], will result in decreased K over time without necessarily resulting in cracking. Thus, only rising K testing was performed in the present work, although it is acknowledged that falling K testing might be appropriate for displacement-controlled situations. Two types of test were performed, one based on applying incrementally increasing K to a single specimen until failure. The other used a single load application per specimen. Whilst the former requires fewer tests, it may result in crack blunting due to sustained load strain before the specimen K exceeds K _ISLC , thereby giving misleading results. On this basis, it was considered appropriate to examine both methods.

The present article, therefore, presents the results of a programme of work aimed at generating threshold fracture stress intensity factors (K _ISLC ) for parent metal, metal inert gas (MIG) and keyhole plasma arc weldments in Ti-6Al-4V alloys, using two different rising K test methods.

Nomenclature

Experimental procedure

Materials

Various batches of Ti-6Al-4V material were employed as shown in Table 1. These included forgings (M1 and M2), pipe (M3) and sheet (M4 and M5) parent materials, and a MIG weld (M4-MIG) in sheet M4 and a keyhole plasma weld (M5-PAW) in sheet M5. Welding procedures are not shown, but conformed with conventional practice for titanium welds. Some samples of the MIG weldment were postweld heat treated at 650°C for one hour.

Table 1 Material composition compared with relevant standards

Test method

Full thickness, Bx2B (where B corresponds to through-wall thickness) single edge notched-bend (SENB) fracture mechanics specimens were manufactured and prepared, according to BS 7448:Part 1:1991 for parent metal studies and BS 7448:Part 2:1997, for weld metal and HAZ studies. Prior to notching in either the HAZ or at the weld metal centreline, each specimen was etched in an aqueous solution of 10%HNO3 + 2%HF to reveal the different regions of the weldment, and enable accurate notch positioning in the through-wall thickness direction. Fracture toughness tests were performed at 4°C as shown in Table 2.

Two types of SLC testing were performed. Both were based on conventional pre-cracked SENB specimens, prepared as above. Estimates were made of the crack depth, from the measured crack length on the sides of the specimens and multiplying the average of these two values by 1.08, to allow for the effect of crack bowing. Loads were calculated that would achieve appropriate approximate stress intensity factors for the tests.

One test method (step-loading) involved loading the samples in a servo-hydraulic test machine and raising the load in incremental steps with dwells at set load points. Tests started by loading samples to a load resulting in approximately 40% of K at maximum load (as measured from the fracture toughness tests) and continued with step-load increments corresponding to 5% of this value. The displacement of a clip gauge attached across the crack mouth was monitored and the test was terminated once tearing ensued. These tests were performed at 4°C. The test temperature was selected on the basis that the risk of SLC is greatest close to this temperature ^[1] . Post-test assessments included fractographic examination of the specimens to determine the depth of any SLC, and examination of the clip gauge displacement-time trace, for evidence of the onset of SLC. These data were employed to calculate a value for the stress intensity factor at which SLC cracking initiated, K _init , and an estimated value for K _max , determined from the load at the onset of tearing and the combined length of the fatigue pre-crack and the SLC crack. Initiation of SLC in step-loaded samples was determined by observation of the slope of the clip gauge displacement versus time plots, being careful to differentiate between initial primary sustained load strain and SLC (an example is given later in the results section). Crack initiation was often noted in the loading step immediately prior to that which resulted in failure. A threshold K value, K _th , for which no SLC was observed, was calculated from the load prior to that at which SLC was first noted.

The second test method (single-load testing) involved loading a series of specimens to set loads in static loading frames at 4°C. Single-load testing was carried out in cantilever arm, dead weight loading frames with a calibrated loading ratio. Each specimen was loaded under bending to a specific value of K, chosen after consideration of the results of the step-loaded tests. Discrepancies between target and actual K values occurred because of the uncertainty in the shape of the fatigue pre-crack. The specimens were carefully loaded over a period of one minute in an attempt to achieve smooth load transition and avoid dynamic loading effects that might blunt the crack-tip or result in premature sample failure. Testing was continued either until failure, noting the time to failure, or until a period of 720 hours had elapsed. On completion of testing, all specimens were broken open and the fracture faces examined using light and scanning electron microscopy, allowing post-fracture measurements of the fatigue pre-crack dimensions and the depth of any SLC, to permit calculation of the appropriate K values. These data were employed to estimate a value for the stress intensity factor below which no SLC was observed, K _ISLC .

The matrix of tests is shown in Table 2. In most cases, duplicate tests were performed and an average taken, except for the single load SLC tests, for which 9 tests were performed.

Table 2 Matrix of Tests

Results

Microstructure, Composition and Toughness

The microstructures of the parent materials and weldments are shown in Fig 1-2 and the chemical compositions of the alloys are given in Table 1. The materials employed were within the desired range of compositions and microstructures, which enabled assessment of SLC phenomena for different material conditions. The aluminium content of parent material M1 (7.42%wt) exceeded the maximum permitted (6.75%wt) for ASTM Grade 5, but this was deemed acceptable for the purpose of establishing the influence of high Al content. A summary of parent material product form, microstructure and toughness is given in Table 3. Toughness is shown as KJmax, calculated from J at maximum load.

Table 3 Summary of Microstructure

Materials M1 and M2 were mill-annealed forged parts, nominally to ASTM Grade 5. M1 exhibited a coarse microstructure containing a mixture of large, equiaxed and elongated alpha grains and isolated grain boundary beta ( Fig.1a). M2, however, exhibited a mixture of fine equiaxed alpha grains and alpha/beta colonies ( Fig.1b). The oxygen content was quite high (0.18%), but still within specification. Material M3 was a rotary-pierced, beta-annealed pipe, to ASTM Grade 23. The prior beta grain size was quite coarse and contained a fully transformed, Widmanstätten alpha/beta microstructure ( Fig.1c). The aluminium and oxygen contents were low (6.03 and 0.095%, respectively). Not surprisingly this material exhibited the greatest toughness.

Materials M4 and M5 were both mill-annealed sheet, to ASTM Grade 5. The microstructure of M4 comprised a mixture of equiaxed alpha grains and globular grain boundary beta ( Fig.1d). The oxygen content was quite high (0.17%), but still within specification. This combination of microstructure and chemistry gave the worst toughness. The microstructure of M5 comprised a mixture of equiaxed, alpha grains and alpha/beta colonies ( Fig.1e). The oxygen content was quite high (0.16%), but still within specification. Toughness (as estimated from the SLC tests) was clearly significantly greater than that of M4, predominantly as a consequence of microstructure.

The microstructure of the weld metals of the MIG weld, M4-MIG, and the plasma arc weld, M5-PAW, were similar, comprising transformed-beta products in coarse, prior beta grains ( Fig.2). The oxygen content of the MIG weld metal was slightly lower than that of the parent metal, whilst that of the plasma arc weld was significantly greater, but, as expected for titanium alloy weldments, this had little visible influence on microstructure. The microstructures of the HAZs of MIG weld M4-MIG and plasma arc weld M5-PAW were also similar, comprising transformed-beta products, with increasing prior beta grain size closest to the fusion boundary ( Fig.2).

Sustained Load Testing (Step-Loading)

The form of the clip gauge displacement versus time traces is shown in Fig.3, for a specimen from the MIG weld metal. This graph shows, schematically, the stages of the test and the difference between plastic deformation due to primary sustained load strain ( Fig.3a) and crack growth due to SLC ( Fig.3b). Micrographs of the fracture faces of one of the specimens are shown in Fig.4. The deduced values of K _init and K _th are listed in Table 4 for each specimen (both shown as ratios against K _max , but K _th is also shown as an absolute value). Summaries of K _init /K _max , K _th /K _max and K _th are presented in Table 5, for the parent metals.

4a) overview - fracture face

4b) as above but at greater magnification. Note the mixed ductile and 'brittle' fracture morphology

Table 4 Summary of step-loading testing of MIG and PAW weld specimens in Ti-6Al-4V plate.

Note: K _init is the lowest stress intensity factor at which SLC was observed and K _th corresponds to the stress intensity factor below which SLC was not observed.

It can be seen that mill-annealed, forged Ti-6Al-4V alloys with high aluminium (M1) or oxygen (M2) content exhibited values of K _init that were 34-37% lower than K _max during step-loading tests ( Table 5). The sustained load performance of the beta-annealed ELI Ti-6Al-4V pipe (M3) was better, showing a value of K _init that was 27% lower than K _max and a K _th value of 67.1 MPa √m.

Table 5 Summary of step-loading test data for parent materials.

Note: K _init is the lowest stress intensity factor at which SLC was observed and K _th corresponds to the stress intensity factor below which SLC was not observed.

The mill-annealed Ti-6Al-4V plate with globular grain boundary beta phase (material M4) exhibited the worst SLC performance in both as-welded and PWHT conditions. A K _th value of 25.6 MPa √m was calculated, which corresponds to a 30% decrease compared with K _max .

The fusion and heat affected zones of the MIG weld (in material M4) exhibited better SLC performance than the parent metal, both in as-welded and PWHT conditions. The minimum K _th values for the weld and HAZ were 50.8MPa √m and 57.3MPa √m, in the as-welded condition, and 60.2MPa √m and 65.0MPa √m in the PWHT condition, respectively. The increase in SLC threshold values in the HAZ and weld metal after PWHT (650°C, 1h) signifies the positive effect of heat treatment at a typical stress-relief temperature.

The SLC performance of the plasma weld metal was lower than that of the parent material (material M5) in the as-welded condition. The minimum K _th value for the weld metal was 40.0MPa √m, primarily as a consequence of higher weld metal oxygen content.

Sustained load testing (single loading)

Samples from the plasma arc weld, notched at the weld metal centreline, were loaded to a single load in static loading rigs, either failed soon after loading or remained intact with no crack extension for a period exceeding 720hours. Both the initial K, Ktest, and a value of K, K _max , at which SLC stopped and tearing began, were calculated. K _max was determined in two ways, depending on whether or not the specimens had failed. For the specimens which failed, an extended crack length, measured on broken samples using SEM fractography to determine the amount of SLC, was employed. One test specimen, failed at K=36.7MPa √m, and three others failed at K values of 40.7, 44.4 and 47.4MPa √m, respectively. The other five specimens did not fail, with K values equal to or greater than 41.2MPa √m. Fractographic examination of the specimens, which failed shortly after loading, showed evidence of mixed brittle and ductile fracture morphology, within a few hundred microns of the fatigue pre-crack tip. There were some regions of fluted fracture features that have been associated with either SLC or slow strain rate failure ^[2] .

Discussion

Parent metals

The SLC test results from the parent materials followed the trends predicted by the present authors previously ^[1] with estimated threshold stress intensity factors for SLC being greatest for beta-annealed parent metal microstructures with low interstitial element contents. In absolute terms, the mill-annealed parent metal performed worse than the beta-annealed pipe, but of the mill-annealed parent metals, the two forgings appeared to perform better than sheet material, despite a chemical composition and microstructure that might be anticipated to increase susceptibility. This trend may be due to greater basal plane texture in the sheet, resulting in greater susceptibility to SLC and reduced toughness. The parent metals exhibited K _th values of between 25.6 and 67.1MPa √m, with increasing values in the sequence material M4, M5, M1, M2, M3.

A range of K _th /K _max values was obtained between 0.5 and 0.8. The lowest figure was for the high Al content mill-annealed forging and the highest figure was for the mill-annealed sheet with a more favourable microstructure (material M5). Thus, in terms of K _th /K _max , the parent metals were ranked in order of ascending material identity number (M x, where x=1-5).

It is clear that a wide range of microstructure, toughness and SLC performance is possible for material supplied in the 'mill-annealed' condition. Unfortunately, whilst many aerospace Ti-6Al-4V specifications stipulate microstructure, such guidance is not employed typically for more general engineering applications. Thus, end users buying 'ASTM Grade 5' titanium alloy could be supplied with material that could readily have both an unfavourable microstructure and disadvantageous chemistry. Material sourced directly from a material producer would tend to be tailored to the clients needs but material, bought from a stockist, is unlikely to be as controlled, unless additional requirements are specified by the end-user. From the present work, microstructure and oxygen content seemed to have the greatest influence on SLC performance, thus these two aspects should be viewed as being critical for structures that see high sustained loads.

Weldments

The results from the present study suggest that, for poor toughness parent metal, the weld zone (HAZ and weld metal) tend to exhibit better SLC performance than the parent metal as a consequence of their fully-transformed microstructure. A caveat should be placed on this assessment, namely that no contamination of the weld metal occurs during welding. The as-welded MIG weld metal and HAZ exhibited K _th >45MPa √m compared with the parent metals K _th =25.6MPa √m.

Postweld heat treatment (stress-relief) showed a consistent trend of increasing K _th to a value greater than the equivalent as-welded values. After PWHT, the minimum K _th in the weld zone was 56.3MPa √m. It is believed that this increase in K _th may be due to the effect of microstructural effects, but the exact mechanism is not clear. It is speculated that partial transformation of grain boundary beta to alpha, as well as tempering of pre-existing alpha, results in better SLC performance. Even so, this is an important result, since it indicates that stress-relief should not prove harmful to SLC performance (provided that no surface oxidation or alpha case results from poorly-controlled heat treatment).

The plasma arc weld exhibited lower K _th than the parent metal, but it is noted that the oxygen content of the keyhole plasma weld (0.19%) was close to the maximum permissible for ASTM Grade 5 (0.2%) and significantly greater than that in the neighbouring parent metal (0.16%). The weld metal exhibited a beta-transformed microstructure that partially offset the influence of the oxygen pick-up, but the higher oxygen content was the main factor in the degraded weld metal performance and highlights the need to ensure good gas shielding during welding.

Test method

Step-loading tests were most economical in terms of test specimens required, but the step-loading and single-loading test data did not fit neatly together. The conservative value of K _th , established by step-loading of the plasma arc weld metal, proved to be greater than the lowest K value that resulted in failure on single-load testing. The step-loading test suggested a value of 40.0MPa √m for K _ISLC would be conservative. However, a single-loading test failed at <36.7MPa √m. The failure at K=36.7MPa √m occurred very quickly, whilst five test specimens loaded at K=41.2-44.2MPa √m did not fail, even after 720h. All other failures occurred above 40.0MPa √m.

The lower value of K _ISLC estimated for plasma arc weld metal by single-load testing compared to step-load testing, could be taken to indicate that the single-loading test is more conservative. However, insufficient tests have been made to determine the relative degrees of scatter inherent in the two test methods. Regardless, it is clear that more than two specimens should be tested for the step loading method. Important differences between the test methods include the overall rates of loading and the possibility of crack blunting in the step-loading test. However, the experimental observations made in this work suggested that blunting was not apparent in either fracture toughness or SLC tests, carried out on a range of Ti-6Al-4V materials.

More testing is required in order to confidently state definitive test methods, including the number of step-loading tests that should be conducted to define K _ISLC . From the current work, a summary of preliminary K _ISLC values is presented in Table 6 for the materials investigated. It is noted that K _ISLC will be influenced by test specimen thickness, and it is recommended that test specimens be taken that are representative of the thickness of the component for which an SLC threshold is to be determined.

Table 6 K _ISLC , based on step-loading tests

One SLC testing strategy that was not explored in the present programme was that attempted by Takatori et al. ^[3] , who investigated the influence of displacement rate during fracture toughness testing on KIC. Their results for mill and beta-annealed high aluminium and oxygen material show that slower displacement rates result in lower KQ values, but further investigation would be necessary to determine if these values were representative K _ISLC parameters.

Practical implications

If toughness and SLC performance are important in a particular design, supplementary specification requirements should state that parent material should not exhibit a microstructure comprising equiaxed alpha grains and globular grain boundary beta phase. Aluminium content should also be close to 6% and oxygen content less than or equal to 0.17%. This compares with maxima of 6.75 and 0.2%, respectively, for ASTM grade 5. In the present investigation, material that met these requirements (material M2, 3 and 5) exhibited K _th greater than 45MPa √m.

When SLC testing is carried out for a particular application, the thickness of the test specimens should be comparable to the thickness employed. During fabrication of Ti-6Al-4V, low oxygen content consumables should be employed and care should be taken to avoid oxygen pick-up. In order to minimise oxygen pick-up, procedure qualification and welder qualification test coupons should be analysed for oxygen in addition to applying more conventional weld colour assessments. Stress-relief heat treatment appears not to degrade SLC performance, provided no oxygen contamination occurs as a result. Indeed, lowered residual stresses should make SLC crack initiation or growth less likely.

Conclusions

The threshold stress intensity factor for sustained load cracking, K _ISLC , in Ti-6Al-4V parent metals and MIG and keyhole plasma weldments in the as-welded and postweld heat treated conditions was assessed. Step-loading and single-load testing techniques of through-thickness SENB specimens, notched in the appropriate zones, were employed for this purpose. The following conclusions are drawn from this investigation:

1. Parent material with a beta-annealed (fully-transformed) microstructure and low aluminium and oxygen contents proved to exhibit the greatest K _ISLC value, 67.1MPa √m. Parent material with an equiaxed alpha and globular grain boundary beta microstructure and 0.17% oxygen had the worst SLC performance, K _ISLC =25.6MPa √m.
2. Forged material proved to have K _ISLC values similar to sheet, at least for the two batches examined in the present work.
3. The SLC performance of weld metal and HAZs is probably greater than or comparable to the parent metal, provided that the oxygen content of the weld metal does not exceed that of the parent metal.
4. MIG weld metal and HAZ with 0.16 and 0.17% oxygen, respectively, exhibited K _ISLC values greater than 45MPa √m in the as-welded condition.
5. Keyhole plasma arc weld metal, with 0.19% oxygen, exhibited K _ISLC <36.7MPa √m, highlighting that, despite a beta-transformed microstructure, weld metal performance can be poor, if oxygen content is too great.
6. Stress-relief heat treatment at 650°C for 1h had a small beneficial effect on SLC resistance of both WM and HAZ.
7. Step-loading has proved to be a useful way of studying SLC. However, the technique needs further development for the generation of quantitative data, before it can be recommended as an alternative to single load testing.

Acknowledgements

The work was funded by Industrial Members of TWI, as part of the Core Research Programme. Particular thanks go to Permascand AB for contributing the Ti-6Al-4V plasma weldment.

References

1. Kostrivas T., Smith L. and Gittos M., December 2003, 'Review of sustained load cracking phenomena in titanium alloys under static and dynamic loading', TWI Member's report 13842.01/03/1147.2.
2. Nixon R C C and Hawkins D N, March 1999, 'Nature of fluted fracture observed in welds in titanium plate', Materials Science and Technology, Vol.5, pp.288-292.
3. Takatori H., Chiba Y. and Ogura T., May 1992 'Effect of microstructure on sustained load cracking behavior of Ti-6Al-4V alloy', Tetsu to Hagane, Vol.78(5), pp.837-844.