AFGROW

Information and insight from fracture mechanics experts

Using the AFGROW Growth Increment Preference

The capabilities available in AFGROW Predict, Preferences menu are often missed by many casual AFGROW users. One very important preference is the growth increment option. The growth increment controls how often AFGROW will recalculate the normalized stress intensity solution (beta value) as a function of the amount of crack growth since the previous calculation. There is a trade-off between the accuracy of the prediction and the total run time. For cases in which crack extension per cycle is very small (say, 1e-6 in.), the change in the beta-value will be minimal (for most cases). For this reason, computer run times can be substantially reduced by taking advantage of the growth increment option.

The default growth increment is set to be 5% of the crack length since the last calculation. This means that the growth increment for each calculation will increase with the crack size. Since most of the calculated life is expended at the shorter crack lengths, this is a reasonable approach. Many users may not realize that the output beta (and associated K) values depend on the growth increment selected. If a user desires the "exact" beta (or K) for each cycle, then it is important to select the appropriate cycle x cycle growth increment. Please refer to the User's Guide or on-line help for more information.

 For example:

Advanced Model (Double Corner Cracked Hole)

Plate Width = 3.25 in., Thickness = 0.25 in., Hole Dia. = 0.15 in., Hole Offset = 1.625 in.

C = 0.0534, A = 0.0655

Constant Amplitude Loading, Smax = 20, R = 0 

The following run times resulted as a function of growth increment (shown for AFGROW Version 4.11.14 and 4.12.15):

Note: The decrease in run times for the new version of AFGROW is attributed to the faster CPU used. Also, the growth increment of 0.0 is the cycle x cycle option. 

As can be seen, there can be a very substantial run time penalty for running a cycle x cycle prediction. However, the increase in prediction accuracy may not warrant the increase in run time. This, of course, can vary from case to case. If a low growth increment is resulting in long run times, it may be helpful to experiment with a larger increment.

The bottom line is that users should be aware of this and use a growth increment that is suited to a given problem.

 

Comments are closed
!-- Google tag (gtag.js) -->