Lower-Bound Thresholds for Fatigue-Crack Propagation
under High-Cycle Fatigue Conditions in Ti-6Al-4V

 

Mr. B. L. Boyce* and Prof. R. O. Ritchie
Department of Materials Science and Mineral Engineering
University of California at Berkeley
Berkeley, California 94720
 

1. Introduction

The high cycle fatigue (HCF) problem experienced in military turbine engines involves ultra-high frequency (>1 kHz) cyclic loading of the turbine components at very high mean stresses. Because the frequencies are so high, cracks can very quickly grow to failure even at low growth rates. For example, a crack propagation rate of 10-10 m/cycle results in 2 mm of growth after only 3 hours. This catastrophically fast propagation necessitates that component service conditions are based on some threshold for fatigue crack propagation (FCP). A major thrust of this program is to identify critical levels of microstructural damage under which flaws can or cannot grow under representative HCF conditions. 

Because in-service failures originate from small flaws (i.e. regions of foreign object damage, fretting, etc.), the most experimentally relevant studies will identify small crack FCP thresholds whose crack sizes are in the range of a few hundred microns. Long crack thresholds generally do not represent the lower-bound for small crack propagation (Figure 1). There are three common phenomenon which explain why small cracks can grow higher growth rates than long cracks [1]:

For example, long cracks in titanium alloys are shielded by roughness-induced crack closure in the crack wake [2,3]. This results in an effective reduction in the applied cyclic stress intensity, D Keff = Kmax - Kclosure. However, small cracks with little or no wake do not experience closure effects and thereby experience the entire applied D K.

Since it is unlikely that future industrial screening processes will involve characterization of small crack behavior, it would be beneficial to identify modified long-crack tests which exhibit behavior similar to small cracks. There are several techniques available to mimic some characteristics of small-crack behavior in long-crack tests. The long-crack tests to be compared include:

1. High Mean Load Tests. By testing at high mean-loads, the minimum stress intensity, Kmin, is above the stress intensity below which closure occurs, Kclosure, thereby minimizing or eliminating extrinsic crack tip shielding [3,4].

2. Constant Kmax Tests. Rather than testing at a constant load-ratio, a constant Kmax test allows the mean to increase as the cyclic stress intensity approaches threshold. This facilitates exceptionally high mean loads at threshold which would not be possible in a constant load-ratio test, thereby minimizing the effects of crack closure [5-8].

3. Compression Precracking. When a far-field compressive overload is applied to a crack, a residual zone of tensile stress is left ahead of the crack tip. Compressive cycling can then be used to propagate this crack through the residual overload zone. The crack arrests at the edge of the residual zone leaving minimal damage ahead of the crack tip. The crack wake is then sectioned off resulting in a short crack (~100 m m long) with no compressive residual stress ahead of the tip [9].

4. Razor Micronotching. Typical machined notches are much too large compared to microstructural dimensions to approximate the behavior of sharp cracks. Precracking is traditionally used to grow a crack from the notch to a length sufficient to eliminate any notch effects on the stress distribution. However, by using a razor micronotching technique, the root radius of the notch is smaller than the characteristic microstructural dimensions. The stress distribution from the micronotch is sufficiently similar to a crack that propagation data can be collected right from the notch. The micronotch has the advantage of having little or no residual zone of damage ahead of the notch tip, and no wake.

 While conventional fatigue tests are conducted at relatively low frequency (1-50 Hz), an important component of this program is fatigue testing at frequencies representative of in-service HCF conditions (~1 kHz). When frequency effects are observed, they are generally attributed to environmental factors, strain rate dependency, and/or creep deformation at the crack tip.

2. Procedure

The primary material of interest in the early stages of the HCF program is Ti-6Al-4V. To facilitate comparison of results, all research programs are studying a specific microstructure specified and supplied by the Air Force. This alloy is in the solution-treated-and-over-aged (STOA) condition: a bimodal microstructure with lamellar a and b lath colonies in a matrix of approximately 50 vol% primary a (Figure 2). Compact-tension (C(T)) specimens were sectioned from the supplied slabs in the L-T orientation.

Fatigue crack propagation tests were conducted according to ASTM E647. Crack length was measured via the back-face strain compliance technique as well as periodic optical verifications. During all tests, load shedding was performed using a constant K-gradient, C = -0.08 mm-1:

D Kfinal = D Kinitial exp [ C * (afinal - ainitial) ]

Fatigue crack propagation tests were performed at 50, 200, and 1000 Hz. The 1 kHz test were performed using a MTS load frame modified with (1) a 3-stage voice-coil servovalve, (2) an acceleration compensated load cell, and (3) a stiffened frame for minimized resonance [10].

3. Results

Figure 2 shows low frequency (50 Hz), R = 0.1 fatigue crack propagation data. A total of 5 specimens were used to obtain this data and the overlap indicates that there was very little specimen-to-specimen scatter. Furthermore, the data were found to correlate well with previous results on a similar microstructure [11].

The effect of mean load on fatigue crack growth is presented in Figure 3. Three load ratios (R = 0.1, 0.5, and 0.8) were tested. Crack closure was monitored as a deviation from linearity in the unloading compliance curve. Of the three load ratios, crack closure was observed only at R = 0.1. The closure stress intensity, Kcl, was found to be 2 MPaÖ m independent of cyclic stress intensity, D K (Figure 4). This is indicative of (roughness induced) crack closure as reported previously in titanium alloys [2,3]. The FCP threshold, D Kth, was 2.9 MPaÖ m for both R=0.5 and R=0.8. This high-R threshold differed significantly from the 4.6 MPaÖ m threshold measured for R = 0.1. The mean-load dependence of threshold is presented in Figure 5. This result also supports the conclusion of a closure value D Kcl ~ 2-3 MPaÖ m.

High frequency growth rate data, i.e. 200 Hz and 1000 Hz, are shown in Figures 6 and 7. The results correlate well with 50 Hz behavior in the intermediate D K regime. 1000 Hz threshold behavior appears to overlap 50 Hz data at R = 0.1. However, at R=0.8, the 1000 Hz data deviates from 50 Hz data near threshold. The 50 Hz threshold is at D K = 2.9 MPaÖ m whereas the 1000 Hz threshold is at D K = 2.5 MPaÖ m. Further testing is required to determine if this difference is statistically significant.

4. Discussion.

This report summarizes data collected to date. Detailed analysis of these results will commence when the initial series of data collection is closer to completion. However, there are some salient conclusions which can be stated from the partial set of data contained in this report.

1. The lowest bound threshold data were obtained at 1000 Hz at R = 0.8. This threshold at D Kth = 2.5 MPaÖ m was observed at growth rates as low as 10-11 m/cycle.

2. In the 50 Hz FCP data, at load ratios as low as R = 0.5, extrinsic crack tip shielding mechanisms appear to be eliminated resulting in a lower-bound threshold of 2.9 MPaÖ m.

3. 1000 and 50 Hz FCP data are nearly identical with the possible exception of small deviations in threshold in the 1000 Hz data.

4. Crack closure appears to remain constant at Kclosure = 2 MPaÖ m, between D K = 6 MPaÖ m and D K = 13 MPaÖ m.
 

 Acknowledgements

This work was funded by the Air Force Office of Scientific Research under the Multidisciplinary University Research Initiative on "High Cycle Fatigue", at the University of California at Berkeley, Grant No. F49620-96-1-0478.

References

1. R. O. Ritchie, "Small Cracks and High-Cycle Fatigue," in Proceedings of the ASME Aerospace Division, J.C.I. Chang et al., eds., ASME, New York, NY, 1996, p.321.
2. M. A. Hicks, R. H. Jeal, and C. J. Beevers, Fat. Engin. Mater. Struct. 6, 51 (1983).
3. M. D. Halliday and C. J. Beevers, J. Test. Eval. 9, 195 (1981).
4. K. T. Venkateswara Rao and R. O. Ritchie, Int. Mater. Rev. 37, 153 (1992).
5. H. Döker, V. Bachmann, and G. Marci, in Fatigue Thresholds: Fundamentals and Engineering Applications, J.    Bäcklund et al., eds., EMAS, Warley, UK, 1982, p.45.
6. S. Mall, J. A. Perez, and T. Nicholas, Eng. Fract. Mech. 37, 15 (1990).
7. G. C. Salivar and F. K. Haake, Eng. Fract. Mech. 37, 505 (1990).
8. G. Marci, Eng. Fract. Mech. 41, 367 (1992).
9. T. Christman and S. Suresh, Eng. Fract. Mech. 23, 953 (1986).
10. J. M. Morgan and W. W. Milligan, in High Cycle Fatigue of Structural Materials, Paul. C. Paris Symposium, W. O. Soboyejo and T. S. Srivatsan, eds., TMS, Warrendale, PA, 1997, in review.
11. A. W. Thompson, J. C. Williams, J. D. Frandsen, and J. C. Williams, in Titanium and Titanium Alloys. Scientific and Technological Aspects, J. C. Williams and A. F. Belov, eds., Plenum, New York, NY, 1982, p. 691.

 

Figure 1. A typical comparison between long crack data and small crack data. Note that the small cracks can propagate at cyclic stress-intensities below D Kth.

 

Figure 2. Baseline fatigue crack propagation data for STOA Ti-6Al-4V.

 

 

 

Figure 3. The effect of mean load. R=0.5 data appears to coincide with R=0.8 data near threshold.

 

 

Figure 4. Closure stress intensity, D Kcl , as a function of cyclic stress intensity, D K.
 

 

Figure 5. Mean load dependence on threshold values, D Kth and Kmax, th.

 

Figure 6. 200 Hz fatigue crack propagation data compared to conventional data.
 

                     
 
Figure 7. 1000 Hz fatigue crack propagation data compared to conventional data.