cycles at R =0.9, k'' cycles at R = 0.7, and in each individual sequence, 1 cycle of R = 0.23, this cycle being formed by the transition between the other two R ratios. This cycle had a starting ∆K of 11.6 MPa m1/2. Crack lengths were measured to an accuracy of ±14 µm with a high precision electrical potential system described in [7]. Tests were conducted at between 8 and 20 Hz on a digitally controlled servo hydraulic machine which controlled test loads to better than 1 % of the maximum load applied during the test. All tests were carried out in a laboratory air environment at between 15-25°C. n´ Load R* R=0.0 Time n Pmax Load R=0.9 R=0.7 Time k k´´ Pmax b) SVAL-2 a) SVAL-1 Figure 1: Definition of two different spectrums, SVAL-1 and SVAL-2 composed of small cycles in high R and underloads. Crack length and cycles data were recorded automatically throughout the application of the test sequences. Crack growth rates da/dN were calculated by determining the total number of cycles of all ranges required to grow the crack successive increments of 0.1 mm. Thus the calculated da/dN was an average contribution from all types of cycle. Data were represented as crack growth increment (a) Vs total sequence cycles (N), and also as da/dN Vs ∆K. An acceleration factor γ, was defined as follows:- ( ) ∆ data growth crack amplitude the constant of summation by linear calculated increment length crack final ( ) ∆ in test measured increment length crack final γ f t a a = The linear summation calculation required crack growth integration, and was performed by the AFGROW crack growth software model [8], with no load interaction effects. The model was provided with CAL data for Ti-10-2-3, which had been measured at R = 0.1, 0.7 and 0.9. The data were curve fit, and the fitted curves used to calculate the crack growth increment produced by the completed sequences, described above. It was established that the curve fits represented the original data to an excellent accuracy in the near threshold crack growth rate regions of interest (Figure 2). The excellent agreement between experimental data for R = 0.4 and predicted line for 0.4 may be observed. The line on the extreme right of the data is that predicted for R = 0. RESULTS It was found that under the SVAL-1 spectra, acceleration effects were observed only for R* values of 0.9. No acceleration effects were found under any of the SVAL-1 spectra for R* =0.7. As the number of underloads in a sequence at R* = 0.7 increased from 1 to 50 the average da/dN increased from 10-11 to 10-10 m/cycle. However, this was a consequence of the increases in the numbers of damaging cycles and no acceleration relative to a linear summation was observed. At R* = 0.9, significant accelerations were found relative to linear summation with γ values from 1.5 to 2.2 (table 2). Fatigue crack increments Vs cycles and crack growth rates can be seen in figure 3 for these tests. ∆K in figure 3 refers to the ∆K of the small cycles in the sequence. It can be seen that the value of the ratio n/n', changes the observed growth rate. For example in
RkJQdWJsaXNoZXIy MjM0NDE=