How to Install AFGROW

[Chronos516]

I know this is probably not the best question to ask here, but I am having problems trying to install AFGROW on my computer. I have installed both the full version and the light version on my computer. When I tried to save a .dax in AFGROW, the program crashes and tells me there is microsoft visual C++ runtime error. The application has requested the Runtime to terminate it in an unusual way. This is not really a engineering technical question but just wondering if anybody can help. Do I need to buy a license key or something to get this to work properly? I thought AFGROW was still free.

 

[alexfx]

I think following post on AFGROW blog should give you direction on how to solve your problem.

http://www.afgrow.net/blog/post/Installation-Instruction-MSXML-40-download.aspx

 

[blakmax]

When you finally install AFGROW you need to recognize that the algorithm for crack growth under composite patches is flawed. It assumes that the crack continues to grow under a Paris Law. This is not correct. Stress intensity for a bonded repair approaches an asymptote and never exceeds that asymptote, as discovered by LRF Rose. Hence crack growth after repair is actually independent of crack length. The original work by Mohan Ratwani contains experimental data which clearly demonstrates that his model diverges from his experimetal data which remains linear in most cases.

Use AFGROW if you wish, but your results will be ultra conservative.

Regards

Blakmax

 

[harterja]

I realize that I am a bit biased, but it may be useful to members of this forum to consider some additional points in response to Blakmax's concerns about AFGROW's bonded repair capabilty.

A bonded repair will initially tend to keep K constant, but there is a physical limit to the adhesive shear strength of any repair. Local disbonding will typically occur at the crack tip because of the inherent stress singularity. This has been well documented. As a crack grows, the disbond will generally grow with the crack, and the ability of the repair to bridge the increasing load will eventually fail.

Dr. Ratwani's approach accounts for the adhesive disbond as well as the effect of out-of-plane bending. His model applies a finite matrix to simulate the adhesive interface and stops calculating the beta correction when the crack length reaches 2 inches. It's certainly not perfect, but is based on a physical model.

The result may be on the conservative side, but the over simplified assumption of a constant K is a bit too optimistic in my opinion.

 

[blakmax]

Hi Harterja

I'm still not convinced. The stress singularity is significantly smaller for a repaired crack which will grow as if it is a small crack, and hence with a smaller singularity than prior to repair. Secondly, adhesive fatigue requires high shear strains, and I am not convinced that the shear strains at the tip will be higher than at the centre of the crack.

However, even if we assume you are correct, the disbond produced would be of a consistent size, and hence the effect should still be relatively linear. The probable reason for the discrepancy between Ratwani's predictions and his own experimental results (which are linear) is as you state the removal of beta correction at a nominal crack size. Why? Removing the beta correction may be appropriate for modelling unrepaired cracks, but for repair cases, the crack will still be growing in the same manner as the same small crack, so any correction factor should not be arbitrarily removed. Maybe then his predictions would more accuratley match his experiments.

All of this discussion aside, it still is a valuable tool to provide a conservative estimate of crack growth at least in the short term. The errors only become significant for longer cracks.

Regards

blakmax

 

[harterja]

You seem to have made up your mind, so I wasn't trying to convince you of anything. I simply wanted to give the readers another point of view and let them decide for themselves.

There were two reports documenting the work done at the Air Force Research Lab (Structures Div.) on this subject: WL-TR-97-3105 and WL-TR-97-3107. The first documents the technical approach and the second documents the experimental verification. The C-Scans taken on the test specimens clearly show the disbonds growing with the crack tips, and the crack length vs. cycle data is not linear (as you would see if K were indeed constant).

I want to be clear about a couple of points you made in your response. First, I can only assume that you meant to say that the yield zone size is smaller for a repaired structure. The point I was making about the singularity was that the adhesive tends to fail at the crack tip under cyclic loading, and a disbond forms around it. Also, I never said we removed the beta correction for long cracks. I said that we stopped calculating it. We used the beta correction calculated for the 2 inch crack for all crack lengths greater than 2 inches. If you look at plots of the beta correction vs. crack length, you will see that they seem to be approaching an asymptote.

 

[blakmax]

Hi Jim

Thanks for the clarification. I am happy to defer to your superior knowledge of AFGROW based on your publications. Further, for even medium term predictions, the divergence in results is not worth argueing about. I'll still hold my concerns about the accuracy of longer term predictions with the acknowledgement that the predictions are conservative anyway so there is not an airworthiness issue. Hence, I have no reservations in recommending AFGROW as a design tool because if you can meet your design requirements using these predictions, users will with a high degree of certainty not exceed those growth rates.

After all, we could argue for ages about the finer points of how many angels we can fit on the head of a pin, when most users really want to know if the pin will hold their trousers up!

Thanks for the discussions.

Max