Information and insight from fracture mechanics experts

2014 European AFGROW Workshop Summary

Predict, Preferences Issues/Suggestions Output intervals printed in "hours" if the option to display life in hours is selected in the Output Intervals tab Also, the crack length plots should be converted to hours In the Propagation Limit tab, the option to stop at a cycle limit should be automatically switched to hours if the User has elected to display the life in hours. For the Lug Boundary Conditions tab, it is currently not possible to use the bearing B.C. for through cracks - this needs to be fixed ASAP. Add some type of warning in the lug dialog box to let User's know which B.C.s are being used - or at least notify them that the B.C.s are set in the Predict, Preferences dialog. Consider adding a failure criterion based on R-curve data Issues/Suggestions/Action Items Related to K-Solutions Consider adding an option to control the % of the axial load solution that is used to approximate the out-of-plane bending solution for straight through-the-thickness cracks. Action item for Jim: Compare the current oblique crack solution for the through crack at a hole to the results of differing % of the axial solution for a straight crack at a hole (in terms of the predicted life). Add a notification in the weight function dialog box to explain the limitations of the stress distributions for part-thru cracks. Add the capability to use the current 2-D User-Defined Beta model for 2, interdependent through cracks that can be assigned different plate thickness values. This is needed to accommodate the NASGRO crack growth rate model that is a function of thickness, but would also allow the local stress state to be estimated independently for each crack. It would also be a quick and easy way to solve additional cases without developing a Plug-In module. An additional crack growth model could be added to the Classic Interface showing an image of an "L" section with a crack defined along each leg with 2 thickness parameters. Action for Jim: Compare the results of using the beta correction capability in two separate runs (one for a part-thru crack to transition, and one for the thru crack portion) to a single run using my "optimized" transition method (part-thru --> thru crack). The purpose is to see if this will eliminate the error caused by the addition of integration points to avoid high slope changes when the normalized stress distribution is transitioned to 1.0 in the a-direction for radial distances > the plate thickness. Restore the beta correction capability for advanced through cracks at holes. It was disabled for the current release because of problems caused when two through cracks merge and the number of crack tips change. We will have to also come up with a workable method to address this issue. Note: The ability to use beta corrections for each loading case independently for both compression and tension (6, independent corrections as a function of crack length and a few other secondary parameters was presented. This capability is accessed from the Stress Intensity Filters menu. This is why the output has been expanded to display beta values for both tension and compression. Action item for Jim: Update the User's Guide to make it clear that the current offset correction for the bearing load case assumes the plate is constrained to prevent in-plane bending. General Interest Item Consider adding more warning messages to Users about input parameters than could result in poor life predictions. The concern here is for more novice Users who may be tempted to think that AFGROW's User Friendly Interface is a substitute for a good working knowledge of Fracture Mechanics. This will have to be a continuing effort since it is difficult to foresee all of the possibilities. New Spectrum Tool Add the ability to randomize load levels within a given sub-spectrum. Carefully consider issues related to cycle counting 3-channel load data. The goal of this new capability is obviously to allow the option for the axial, bending, & bearing stress fractions to change for each stress level as may be expected in reality. Northrop as been asking for this capability for some time, and they tell us that they have data ready for use. However, for general use, there are many ways to manage 3-channel spectra. One possible way would be to use a single channel spectrum that has been cycle counted directly from flight test data, and break out the 3 load cases based on some estimate of the load fractions for each level based on the users best information. This would be very tedious, but the use of the XML file format could be used to identify load cases and make it possible to help to automate the process. This type of improvement will have to continue to be a work in progress, and I'm sure there will be other good ideas. In any case, the new tool will be very useful for development of standard, single channel spectra. The XML format will also allow us to post-process AFGROW output files to do things like quantify which sub-spectra are most damaging.

Predict the crack growth life of a structure containing internal through cracks in a row of fastener holes

A short paper by Mr. James Harter discussing the use of AFGROW to predict the crack growth life of a structure containing internal through cracks in a row of fastener holes is now available on the AFGROW web site. The paper provided details on the use of the Advanced Model Interface and provides a direct comparison with FEM results using StressCheck. To obtain this document ( go to the My AFGROW section.

Load Interaction Models - Retardation Models - Something to think about

It goes without saying that load interaction is important if the life prediction is to represent reality and not simply be a go no go measure.  Often times go no go is good enough.  But a lot of the time in the management of aging fleets, a real life prediction can be the difference between retirement or return to service. Retardation can account for 70 percent of the life and that in it self should create wonder!  Well where does this miraculous life enhancement come from?  The neat thing is that nobody really knows.  It is true that few men created (made up) some useful models based upon rational guesses.  These models have been tweaked for years and that’s what makes them so very useful.  They can be tuned to the specific aircraft, usage, and component.  All that’s needed is an attempt at the real spectrum, and the testing equivalent and a test lab.  The number of coupons is left as an exercise.  Of course another avenue is to have service history of cracking and a detail record of how the aircraft was used.  At one time we referred to this as correlating to flying fatigue test articles.  This is because even the OEM has to assume how the customer is going to use the aircraft, and they test to their assumption, reality is sometimes different.  So what?  I suppose any one who has even a cursory knowledge of fracture has come across the notion of the fictional “plastic radius.”  It is fictional because the extent of plasticity at a crack tip is definitely not shaped like a radius, so much for the physics of the problem.  A lot of smoke and mirrors and late night sections have been spent on the subject of the plastic radius in all its manifestations, from plane strain plane stress, fracture toughness coupon requirements, to net section yield as a crack propagates.   It is no surprise that load interaction investigators got hooked on the notion of a plastic interaction zone which increases and/or decreases with spectra level and crack length, some later even attempted account for plasticity in the wake of the crack along the flanks.  Simply amazing rational insights but not yet reality.  The point of this is that current state of load interaction models all rely on the notion of plasticity, some “out and out” some “inferred” but all do.   If one digs into the subject even a little he will quickly realize that the benefit arising from plasticity IS dependent upon constraint.  The assumed closure mechanism that all the load interaction models are based on either at the crack tip or along the flanks as the case may be is totally dependent on there being enough elastic material surrounding the crack to provide sufficient strength to squeeze or constrain the plastic deformed region back into its original (undeformed) shape (volume).   The “how much is needed” is why I started this note.  I don’t know the answer but I know its important and possibly the real mechanism of retardation.  But one thing I do know is that when the crack is approaching a free edge, either the part or another hole, there certainly is not sufficient elastic material surrounding it to create the constraint or closure forces.  Without closure forces there is no retardation.   So you may grab one of the interaction models and do a fantastic job of modeling the cracking scenario and make a perfect prediction but you may be off and in the wrong direction!  Why, all the interaction models assume that there is “enough” elastic material surrounding the crack tip.  There are of course mitigating circumstances such as very little life available due to the steepness of the beta and/or the load level of the spectra. So caution is urged in applying load interaction models carte blanche to the crack growth problems.  Edge ligament cracking is a prefect example of retardation decaying, as are those pesky continuing damage cracks into adjacent holes.  It seems then the only time that the current load interaction models are fully effective is out in the middle of a panel with no discontinuities. And we haven’t discussed relaxation of the plastic zone with time under self weight, perhaps for aircraft that sit on the ramp for weeks between flights. What happens to the load interaction then? I can’t remember anyone really pointing out this subtle nuisance … so I thought I might.

Curve Fit Crack Growth Rate Data

(NASGRO 3 versus NASGRO 4?) The real question that needs to be asked is: where is the “dadN test data” upon which the NASGRO 3 and NASGRO 4 databases are based?  I’ve heard arguments about the differences in the Forman equation for years now and believe we have become blinded to the central issue … BOTH ARE CURVE FITS.  Somehow the community has been lead to believe that curve fit equations are better than honest test data.  The inaccuracy pointed out in ignoring the double knee indicated by the real test data versus curve fit da/dN is far more significant than the threshold issue, which NASGRO 4 supposedly addresses.  Of course it is wrong to ignore either but to make a point. If one would download or the reference located on the AFGROW documentation web page, and go to FIGURE B-4a. NASGRO EQUATION FIT FOR 2024-T3 CLAD AND BARE SHEET, L-T (M2EA11AB1) WITH POSITIVE R VALUES on page B-8, we might be able to understand all the tumult by examining what curve fitting does. For example suppose a spectra which has many R=0 cycles. And consider the following: Delta K 5 10 20 Actual dadN 2.0E-7 6.5E-6 2.0E-7 Curve Fit dadN 3.5E-7 3.5E-6 3.0E-7 Act/CF 0.57 1.85 0.66   The point is: The life prediction at DK = 5 would be 0.57 times shorter than it should be. The life prediction at DK = 10 would be 1.85 times longer than it should be. The life prediction at DK = 20 would be 0.66 times shorter than it should be. Your guess is a good as mine as to how accurate the cumulative life prediction would be but this certainly throws some doubt on it … right? I mean with this kind of accuracy I could just about make up some dadN and be just as accurate.