A. Kostrivas, L. S. Smith, M. F. Gittos
TWI Ltd, Cambridge, UK
Paper presented at the 10th World Conference on Titanium, 13-18 July 2003, Hamburg, Germany
Recent failures of titanium alloy components have drawn attention to sustained load failure phenomena, be they 'cold creep', sustained load cracking (SLC), dwell fatigue or ripple fatigue. These phenomena occur at approximately -50 to 200°C. 'Cold creep' has been used to describe slow extension in titanium under high, sustained loads but a better term is 'sustained load strain' (SLS), since deformation cannot be ascribed to a conventional creep mechanism. SLC is sub-critical crack growth, occurring at stress intensity factors below the fracture toughness. Early service and experimental observations of SLC in titanium were noted in the 1950s but have been largely ignored in non-aerospace fabrications. Dwell fatigue is typified by a square waveform stress cycle incorporating a hold at the high stress. At elevated temperatures this can be considered as creep-fatigue, but dwell fatigue also occurs at lower temperatures (through a combination of SLS/SLC and low cycle fatigue). Ripple fatigue is the synergy of low amplitude, high cycle fatigue crack growth and SLC. In summary, time dependent failures in titanium alloys can occur at stresses below tensile strength (UTS), stress intensity factors below K IC and when dwell or ripple fatigue loads are applied.
In recent years, new applications for titanium have arisen outside the traditional aerospace and chemical industries, the vast majority of which have been successful. However, it must not be forgotten that, aside from the difference between conventional simple static and dynamic performance issues, there is significant difference between the low and high strength grades. In general engineering, ferrous alloy structures are commonly designed following simple rules based upon proof stress or UTS, and/or S-N fatigue curves. This is also quite reasonable for the low strength titanium alloys, which have reasonable ductility and can be considered quite 'forgiving'. However, the high strength titanium alloys are much less forgiving, with common values of fracture toughness less than 60MPam 0.5 and weldment ductility less than 10%. Furthermore, the majority of critical titanium aerospace structures, for which most of the historical data were derived, are designed on the basis of toughness and fatigue crack growth rate (i.e. flaw tolerance). This difference in design philosophy can lead to problems since the service stresses permitted following a general engineering design route may well exceed those deemed sensible following a strategy of flaw tolerance. Similarly, titanium alloy failure mechanisms that are rarely observed in the aerospace sector (where stresses and stress intensity factors are limited to lower values) can be more common in other sectors. Phenomena such as 'cold creep', sustained load cracking (SLC), dwell fatigue and ripple fatigue are amongst those that should be considered when designing with titanium alloys. [1-5]
2 Case histories
An example of SLC, as a consequence of residual stress, was observed in a welding procedure qualification (WPQ) specimen. Here, a 20mm thick Ti-6Al-4V plate had been welded manually using TIG. The weldment had passed penetrant inspection tests shortly after welding, but was found to be cracked after several weeks ( Figure 1). The cracks were located in the parent metal adjacent to the weld, where residual stresses are expected to be high. Evidence of localised strain was observed on the plate surface ( Figure 1b). Cracking is believed to have initiated by sustained load strain (or a combination of the distortion of the plate and SLS), with subsequent propagation occurring by SLC. Fractography showed a faceted fracture morphology with occasional microvoids that is typical of sustained load failure (although the proportion of 'brittle' and ductile fracture can vary) ( Figure 1c).
Fig.1. Photomicrographs showing SLC cracking in a WPQ TIG welded Ti-6Al-4V (ASTM Grade 5) specimen
a) Cross section showing cracking just outside the transformed HAZ
c) Secondary electron image from the crack surface showing faceted morphology
A case was reported by Khaled  of two leaking TIG welded spherical helium tanks due to a combination of sustained load cracking and dwell fatigue. Sustained load cracking here, initiated from a 0.5mm diameter weld pore which acted as a stress raiser in a 3mm thick weld. The welds had passed all the initial inspection requirements and failed after approximately 10 flights. In addition, the oxygen and hydrogen levels in the weld metal exceeded the maximum amounts specified for the Ti-6Al-4V ELI filler wire. It has been reported that increased amounts of oxygen, aluminium and hydrogen decrease the threshold parameter (K ISLC) below which SLC occurs in titanium alloys. 
3 SLC attributes
3.1 Sustained load strain
Sustained load plastic strain occurs in both low and high strength titanium alloys [5-18] and has been observed at constant stresses significantly lower than the ultimate tensile stress and, in some cases, lower than the 0.2% proof stress. Time to failure is dictated by the magnitude of the stress and the form of the time/strain curve is analogous to that of creep with primary (rapid initial strain, with a decaying strain rate), secondary (slow constant strain rate) and tertiary (accelerating strain rate to failure) zones. At low to moderate stresses, only primary strain is observed with no steady state strain occurring. At high stresses, all three zones are observed, but if stresses are less than the 0.2% proof stress, failure might only occur after some decades. Figure 2 shows data from the literature that extend to 24 years for both commercially pure titanium (wrought ASTM grade 2) and cast Ti-6Al-4V. At 85-90% of 0.2% proof stress (180MPa for Grade 2 and 723MPa for Ti-6Al-4V), approximately 3% strain was developed in Ti-6Al-4V and 5-8% strain in commercially pure material, after 24 years. At lower stress levels, significantly lower strains are accumulated, as shown by the shorter duration test data for Ti-6Al-4V and Ti-5Al-2.5Sn. In most engineering applications, SLS will have little influence on performance since the stresses required for failure within 25 years are very high (compared to the 0.2% proof stress). However, for bolted components, allowance should be made for the relaxation of bolt tension that will occur, either by setting a greater initial tension or retightening the bolts after a given period has elapsed.
Fig. 2. Short and long-term room temperature sustained load strain data for a range of titanium alloys. Material C in commercial purity titanium (three batches: a, b & c, each stressed to 180MPa) and material D is castTi-6Al-4V (stressed to 723MPa). The remaining alloys are as shown in the key and were stressed to 80% or 60% (as highlighted) of 0.2% proof stress. [5,6]
3.2 Sustained load cracking
Two principal theories for the incidence of sustained load cracking have been proposed, based on either hydride formation or anisotropy in the alpha phase. [19-24] Even without the precipitation of hydrides, hydrogen may play a role in a manner analogous to that which causes fabrication hydrogen cracking in steels (which itself can be considered a form of SLC). The other mechanism thought to be responsible for SLC is the build up of sustained load strain at interfaces between grains orientated along different differently orientation with respect to the direction of maximum stress. One grain may be orientated such that it undergoes significant deformation, but its neighbour may be orientated such that dislocation propagation or nucleation is unfavourable. The resultant strain accumulation at the grain boundary may result in the brittle fracture of the adjacent grain. This mechanism would explain the SLC fracture appearance, comprising a mixture of ductile and brittle regions. It is believed by the present authors that anisotropy in the alpha phase, a low strain hardening co-efficient and the synergistic effect of hydrogen may all contribute to SLC.
Sustained load cracking appears to be mostly associated with higher strength alloys, which can exhibit a threshold behaviour, such that crack extension will not occur below a critical stress intensity factor (K ISLC). Time to failure can be as long as several days or weeks, just above K ISLC, or as short as a few seconds or minutes, for values closer to the fracture toughness of the alloy. The threshold exhibits complex dependence on temperature, microstructure, composition and initial hydrogen content of the alloy, but, for Ti-6Al-4V grades, values as low as 65% of fracture toughness have been reported ( Fig.3).  Alloys in the mill-annealed condition and containing Al and O at the top end of the permitted levels can exhibit low threshold values, whilst beta-annealed material with lower Al and O contents may exhibit high values. Acicular type microstructures (which bear a complex relationship with prior beta grain size, such that coarse-grained material may have better SLC resistance) provide a tortuous crack path leading to greater threshold values ( Fig.4). Whilst extremely low hydrogen contents, say <1ppm, may decrease susceptibility to SLC, for typical levels within specification for the material grade employed, no strong influence is observed.
Fig. 3. Sustained load cracking data from Yoder et al.  Note that only data from specimens loaded to above K ISLC are shown - PTC-tension test data.
Fig. 4. Effect of microstructure on the stress intensity threshold values for sustained load cracking in titanium alloys 
3.3 Dwell fatigue
Dwell fatigue is particularly relevant to aeroengine components, for which maximum stresses may be sustained over an extended period. Above a particular maximum stress, or stress intensity factor, dwell fatigue life is much shorter than for simple sinusoidal loading ( Fig.5a). It is believed that this is a consequence of the synergistic effect of SLS/SLC and low cycle fatigue leading to earlier crack nucleation and enhanced crack growth rates. At very high dwell stresses (>70% of UTS), dwell fatigue life can be two orders of magnitude shorter than analogous, sinusoidal, low-cycle fatigue.
3.4 Ripple fatigue
Ripple fatigue is applicable to all components that see a reasonably high mean stress, but experience load perturbations from, for example in marine applications, wave loading. Exposure to small amplitude loads (ripple loading) seems to intensify the SLC phenomena, through the combination of SLC and high cycle fatigue. For a given ΔK, it is possible to define a threshold value for mean K, below which ripple fatigue will not occur (K IRLC) ( Fig.5b).  This has been demonstrated to be significantly lower than K ISLC for ΔK set at 5% of K mean. For R=0.9, K IRLC values half that of K ISLC have been demonstrated for Ti-6Al-4V (K IRLC = 25MPam 0.5) and even greater reductions have been noted in alloys such as Ti-15V-3Cr-3Al-3Sn and Ti-15Mo-3Nb-Si.
Fig.5a) Effect of dwell fatigue loading on SLC propensity in titanium alloys expressed by strain accumulation versus applied stress level. For comparison purposes, data for monotonic and sinusoidal loading are also shown(above) 
Fig.5b) Sustained load cracking threshold in titanium alloys under ripple loading.  BA is beta-annealed and degradations are shown as a percentage of K ISLC (above)
Sustained load strain (as opposed to cracking) seems to be universal to titanium and its alloys, [4-24] but its practical/industrial impact would appear to be minimal, given that it will only lead to failure within a reasonable period at stresses well in excess of those that would be expected in components designed effectively using conventional procedures, for example, the application of pressure vessel design codes. Thus, design for flaw-free material under static load can be adequately accomplished by employing codes that disallow maximum stresses greater than 0.2% proof stress.
The authors are not aware of any service cases of sustained load cracking, dwell fatigue or ripple fatigue in commercially pure titanium. Whilst sustained load cracking phenomena may well occur in these grades, it seems likely that they are significantly less susceptible than the high strength alloys. Thus, although no definitive data are available, comfort can be drawn from the many years of successful application of this material in structures designed following ASME requirements. However, sustained load cracking and dwell/ripple fatigue failures have occurred in high strength alloys, such as Ti-6Al-4V. Thus, it is considered prudent to design structures made from these grades employing parameters such as K ISLC or K IRLC (the threshold toughness values below which no SLC occurs) instead of K IC and conventional fatigue limits, where appropriate. For Ti-6Al-4V applications where flaw tolerance is required, it is recommended that extra-low interstitial (ELI) grades (such as ASTM grades 23 and 29) in the beta-annealed condition are employed, with an aluminium content less than, say, 6.4%. However, whilst this recommendation seems reasonable given present knowledge, it must be considered tentative at present.
No published data are available for SLC performance of welds in titanium alloys, but SLC susceptibility of many welds is likely to be low, because weld microstructures often mimic those of beta-annealed parent materials and most welding wire meets ELI specifications. Fabrication should be carried out employing stress-relief to minimise residual stress and good shielding practice is paramount to avoid oxygen/nitrogen pick-up.
Finally, since data on sustained load phenomena are scarce (non-existent in the case of weldments), firm threshold values for design are not available. It is therefore necessary to determine with more certainty appropriate minimum threshold stresses and stress intensity factors for the common titanium grades in both parent metals and weldments. In the first instance, data for parent material and welds in mill-annealed Ti-6Al-4V and beta-annealed Ti-6Al-4V ELI are required to determine appropriate design parameters for these common alloy grades.
- Titanium and its alloys can suffer from time-dependent failures under static loads, or accelerated failure under cyclic loads at typical ambient temperatures. This is manifested in several ways:
- Sustained load strain, analogous in behaviour, although not in mechanism, to creep. This is applicable to low and high strength titanium alloys.
- Sustained load cracking at stress intensity factors below the fracture toughness of the alloy, particularly for high strength alloys.
- Dwell fatigue (low cycle fatigue, incorporating holds at the high stress level), typically relevant to high strength alloys.
- Ripple fatigue (high cycle fatigue at moderate to high mean stress/stress intensity factor), particularly for high strength alloys.
- Since data are scarce for even the most common titanium alloys (and non-existent for weldments), threshold parameter values, below which sustained load phenomena will not occur, cannot be drawn from the literature with anyconfidence. Thus, data and methodologies should be developed which enable designs to take account of sustained load phenomena.
The work was funded by Industrial Members of TWI, as part of the Core Research Programme.
- Burte H M et al, Metal Progress, 67, 1955, 115.
- Evans W J, Bache M R, Fatigue, 16, 1994, 443.
- Pao et al, Scripta Metallurgica et Materialia, 33 (3), 1995, 497.
- Khaled T, J. of Mat. Eng. Perf., 3(1), 1994, 21.
- Drefahl K, Titanium Science and Technology, Proceeding of the Fifth International Conference on Titanium Congress-Center, 4, Munich, 1984, 2387.
- Odegard B C, Thompson A W, Met. Trans., 5, 1974, 1207.
- Meyn D A, Met. Trans., 5, 1974, 2405.
- Smith L S, Gittos M F, TWI Members' report 7321.02/98, 1998.
- Sandoz G, National Association of Corrosion Engineers, 684, 1969.
- Meyn D A, Report of NRL progress, 1971, 11.
- Williams D N, Met. Trans. A, .4, 1973,675.
- 'S4-B explosion', Aviation Week, Space technology 22, 1967.
- Katcher M, Met. Eng. Quarte.,3(8), 1968, 19.
- Miller W H et al, Met. Trans. A, 18, 1987, 1451.
- Imam M A and Gilmore C M, Met. Trans. A, 10, 1979, 419.
- Neeraj T, Acta Materialia, 48 (6), 2000, 1225.
- Hatch et al, Journal of Metals, 2 (1), 1967, 111.
- Suri S et al, Acta Materialia, 47(3), 1999, 1019.
- Pao P S, Wei R P, Proc. of the Fifth International Conference on Titanium Congress-Center, Germany, 1984, 2503.
- Pardee W J and Paton N E, Met. Trans. A, 11, 1980, 1391.
- Howard G et al, Met. Trans., 3, 1972, 469.
- Moody N R, Gerberich W W, Met. Trans. A, 11, 1980, 973.
- Sastry S M L et al, Met. Trans. A, 12, 1981, 8394.
- Rodney R B and Spurr W F, Met. Trans A, 9, 1978, 23.
- Yoder G R et al, NRL Report 7596, Naval Research Laboratory, Washington D.C., 1973.
- Somerday et al, 53 rd NACE Corrosion Conference , paper 267, San Diego, 1998.
- Evans W J, Materials Science A, 243, 1998, 89.
- Pao et al, Scripta Metallurgica et Materialia, 33 (3), 1995, 497.