Practicality of Bonded Joint Analysis 2

[ESPcomposites]

A post from another thread got me curious about elastic-plastic bonded joint analysis, such as those proposed by Hart-Smith.

I have used approaches like this in the research domain or space applications, but have yet to see them translated to the production aircraft side. I suppose as a qualitative study it could be useful, but are these models actually being used on production programs?

If using good design practice (i.e. proper taper ratios, correct joint type), would you really expect the bondline to fail if they have been properly processed? I might expect to see a bond failure due to processing problems, but the analysis cannot predict this. That scenario is the real issue in my eyes, which would make any analysis a theoretical upper limit of capability.

This is not to mention questions about adhesive properties, especially at temperature. In the end, testing and good design practice seems to be the way to go. This may then be combined with a failsafe approach which may assume a processing problem leading to a weak joint. I am not terribly confident in NDI either as poor bonds can go undetected. I have seen this happen on multiple occasions.

So I am wondering what the real world value of a high fidelity bonded joint analysis really is? Is it to just better understand the problem and make better decisions or is structure actually being certified to the analysis?

Brian

http://www.espcomposites.com

 

[SWComposites]

would you really expect the bondline to fail if they have been properly processed?
> with metal adherends, Yes; with composite adherends, No
> with composite adherends, the joint needs to be analyzed using fracture methods, not stress based methods

what the real world value of a high fidelity bonded joint analysis really is?
> in some design cases it can be appropriate and useful

is structure actually being certified to the analysis?
> there is certified bonded structure, using some analysis supported by lots of tests

SW

 

[ESPcomposites]

SW,

Thanks. I forgot to mention that I was specifically interested in composite bonded joints.

I agree that I expect the adherend to fail before the bond, especially in the presence of a notch (or if designing for a hypothetical notch). If so, what is the purpose of of the fracture based bondline analysis since that would not be a critical failure mode?

I suppose there are real conditions like stringers disbonding, where I have seen VCCT used with ABAQUS. Is this what you mean by fracture based models? But I am referring more to the practicality of "standard" type load transfer joints, with models such as those proposed by Hart-Smith.

Brian

http://www.espcomposites.com

 

[SWComposites]

what is the purpose of of the fracture based bondline analysis since that would not be a critical failure mode?
> I meant fracture analysis in the composite adherends

>stringer "disbonding" failures occur in the composite, unless the surface prep is messed up, or there is something wrong with the adhesive.

the practicality of "standard" type load transfer joints, with models such as those proposed by Hart-Smith
> these are not that practical for composite joints, unless
a) one uses composite properties in the analysis,
b) it is a joint design driven by the adhesive, like a complicated step lap joints
And besides, certification usually requires analysis of the joint with disbonds/impact damage, which generally requires a fracture based approach.

SW

 

[MintJulep]

I might expect to see a bond failure due to processing problems, but the analysis cannot predict this

Process failures (i.e. "workmanship or quality problems") are the most common reasons for bonded joint failures.

An analysis that assumes that every joint will be perfect is a fundamentally flawed analysis.

The joint analysis must assess the sensitivity of the joint to potential (and inevitable) process failures, and this must be tied into the FMECA for the assembly. In a production environment it is a certainty that a bond will fail at some point in the service population of bonds. If this failure will result in airplanes falling out of the sky then an alternate joint design is required.

 

[SWComposites]

Yes, and that is why joints must be designed with disbond arrestment features and analyzed with defined flaws.

But, analysis cannot predict the strength of the bond as a function of NDI results.

SW

 

[ESPcomposites]

I guess that is my point. The idealized models by Hart-Smith just don't seem that useful to me because of the real world problems with bonded joints. That is what spawned my question, since another post was about trying to find adhesive properties (which must be nonlinear and are temperature dependent). It's seemingly a lot of work for little value.

I suppose that is why they get far less analytical attention from on a production program than bolted joints (my favorite topic).

SW, but can you elaborate what you mean by "fracture analysis of the composite adherends". I suppose you mean the fracture of the part and not a fracture based failure criterion model. IBOLT, used by Lockheed uses a fracture model, but that is probably not as common as the W-N models (i.e. BJSFM, etc.) which are just semi-empirical criterion.

Brian

http://www.espcomposites.com

 

[SWComposites]

Interlaminar fracture analysis (VCCT type) for delamination onset and propagation using G1c, G2c, mixed mode fracture toughness data. Not the thru-thickness fracture method used by IBOLT.

SW

 

[ESPcomposites]

Maybe I should be more specific when I say "standard load transfer joints". For example, would you suggest a VCCT on a scarf joint? I am only familiar with VCCT use on something like a skin/stringer where the out of plane loads can generate delamination.

For a 30:1 or greater scarf joint, wouldn't the aherend fail before delamination? If not, wouldn't you just increase the scarf angle, based on test, to ensure the delamination failure mode would not occur. Or would that be a poor assumption?

Brian

http://www.espcomposites.com

 

[SWComposites]

Assuming you are talking about bonded repair scarf joints, loaded in tension,

"For a 30:1 or greater scarf joint, wouldn't the aherend fail before delamination?"
> maybe, maybe not, depends on the repair layup,number of overlap plies, repair material, etc, etc,

"wouldn't you just increase the scarf angle, based on test, to ensure the delamination failure mode would not occur?"
>up to a point, and depends on whether you mean delam in the adherend or in the adhesive/interface; you can only increase the scarf length so much, before it becomes absurd, and with longer scarf lengths you will get more waviness in the repair plies, which will reduce strength.

A fracture based analysis may be needed if the scarf joint must be analyzed for flaws and damage, as a stress based approach only works for a pristine bondline

And in simple in-plane loaded lap joints, "peel" stresses cause delams; delams don't just occur do to externally appied loads

 

[berkshire]

""wouldn't you just increase the scarf angle, based on test, to ensure the delamination failure mode would not occur?"

From a practical side of this problem, I have been repairing fiberglass and advanced composite sailplanes since 1974.Most of which originate in Germany. The most common splice joint in glass fiber is a 50/1 scarf joint as specified in the SRM's of the German sailplane manufacturers These same manufactures specify 100/1 for carbon fiber.Getting these bevels without waviness is hard to do and demands a lot of skill and time of the employees. One company in the UK specifies a 30/1 scarf for glass fiber.
Their argument being that it is hard to do a 50/1 scarf, and a properly done 30/1 scarf is as good as an improperly done 50/1 splice. It does not mention an improperly done 30/1 splice. Occasionally a sailplane gets broken again, and in 35 years of repair, I have never seen a repair break again through the old repair. Inevitably the break is just alongside the end of the scarf of the original repair, Never in the break itself.
This may be due to the Germans specifying one extra layer of glass as a "sanding" layer in the repair.
The other point about the failure mode of a fiberglass/ carbon fiber fuselage break ,is that the failure is most often a buckling failure on the compression side. very rarely a tension failure. Wing failures that make it to a repair facility are almost always in compression as a result of hitting something like trees or fence posts.

A tension failure in the structure either at a joint or in the general layup is very rare.
Whilst sailplanes are lightly loaded structures in general, due to the long wingspans and proportionately long fuselage lengths the structures experience very heavy point loads, sometimes greater than advanced composite parts on commercial jet aircraft.
The sailplane industry has been fabricating and repairing plastic aircraft since 1957,and has amassed a lot of knowledge on repairing these structures, however it has take many years for this information to filter into the rest of the aviation industry.
B.E.

 

[ESPcomposites]

This has been a useful discussion.

So at the end of the day, what apparent value are the methods provided in MIL-HDBK-17 (fairly extensive section on bonded joints). I think it provides good knowledge about shear and peel stresses in a bonded joint, which alone has several benefits. But the direct practical applications of them are seemingly limited.

Brian

http://www.espcomposites.com

 

[blakmax]

The RAAF has a complete engineering standard (DEF (AUST) 9005) based on Hart-Smith's work for practical repairs to aircraft structures. It is also the basis of DOT/FAA/AR – TN06/07, Apr 2007 and while it has a focus on metal bonded joints, with a few minor modifications (included in the standard) it is applicable to composites.

Regards

Blakmax

 

[ESPcomposites]

Blakmax, that is interesting. I will have to look into it.

I don't disagree that the methods will generate results that can be useful, provided you have a perfect bondline. But when dealing with composites, you might expect the failure to be in the adherend or due to a poor bondline. In either of these scenarios, the bonded analysis is not adding real value. Therefore, begging the question of it's usefulness.

Limiting the discussion to a scarf joint, I suppose one potential valuable piece of information would be to determine a critical taper ratio (and then validate it by test). This way you could theoretically choose better candidates for test. However, I am not sure if it really works like this, or if candidates are chosen based upon prior know design rules. Perhaps a combination of both would be useful.

Brian

http://www.espcomposites.com

 

[berkshire]

blakmax (Aeronautics)
Didn't you write that paper?
B.E.

 

[blakmax]

Yes, BE, I wrote the standard and two associated handbooks on bonded repair design and application technology. I also wrote DOT/FAA/AR – TN06/07, Apr 2007.

There is a danger that this type of discussion is side-tracked by issues such as "poor bond-lines". I urge people to realise that design can not always account for poor processing. In particular, if the failure is interfacial, then there is a real risk that the interface is degrading and the eventual strength of the bond will be zero, and no design technique can account for that.

Interfacial failures ARE processing deficiencies, and they must be corrected by modifying the processes, not the design. Another factor which is not usually accounted for is micro-voiding in the bondline (which also occurs in composites) due to the materials being exposed to excessive humidity during production. Even damage tolerance analysis can not account for micro-voiding.

It is important to understand Hart-Smith's philosophy for bonded joint design; it does NOT predict bond failure loads. Rather, it makes sure that the strength of the bond is higher than the strength of the adherends, that way the adhesive should NEVER be the locus of failure. Then an adequate overlap length is provided to ensure that the adhesive can achieve that strength. The beauty of this approach for joint design and repairs is that, if the adhesive never fails, then failure will be outside the joint and that can be addressed by existing structural analyses. The design is not dependent on the adhesive performance.

In contrast, according to a survey for the FAA Workshop in Seattle in 2004, about 78% of bonded joint designs in the USA are undertaken using an average shear stress design methodology which was discredited as long back as 1936. The only reason these designs actutally survive is that there are so many conservative factors built into the "design allowable stress" and there are so many tests undertaken to back up this design methodology, that the risk of failure is usually small.

Now, I ask you to consider this: If I design such that the adhesive is NEVER critical by calcualting the adhesive load capacity and comparing that with design loads, then every test I undertake will fail the structure outside the joint. Then why should I do thousands of tests which only measure the strength of the adherends? Surely, if it can be demonstrated that the adhesive is always stronger than the adherends, then the number of tests for certification can be radically reduced, and the cost of certification will also be radically reduced.

The specific problem with bonding to composites is not necessarily the strength of the adhesive, it is the strength of the resin in the composite. If the adhesive is stronger, then failure will occur in the resin between the adhesive and the first ply, or between the first few plies. Another issue is peel stresses, because the peel strength of most resins is also lower than the peel strength of many adhesives. So the same locus of failure can be generated by either shear or peel.

As for scarf joints... Hart-Smith showed that as much as 80% of bond strength comes from plastic behaviour. Scarf joints aim at generating uniform shear stresses. Hence, if the shear stress in a scarf joint exceeds the elastic limit, the joint will uniformly creep and fail under sustained load, so scarf joints can only utilise 20% of the strength of many adhesives. Are they as efficient as people really think? The other limitation is that the scarf typically is about 3 degrees to generate a uniform shear stress. That is about a 30:1 taper. So for a 1/2 inch thick composite wing structure, I must remove 15 inches of perfectly good material each side of the damage just to get a uniform shear stress? I think that if I told my senior engineer that I wanted to carve a circle of at least 30 inches diameter of perfectly good structure away to do a repair, he would seriously question my design capability.

Did I hear a cat being released amongts the pigeoons?

Regards

Blakmax

 

[Compositepro]

Somewhat relevant to this discussion, I've seen lots of testing, like flatwise tensile and peel on sandwich specimens, where the failure is completely in the the core. The results are reported as adhesive properties even though they have nothing to do with the adhesive.

 

[SWComposites]

"even though they have nothing to do with the adhesive"

Not exactly correct - the FWT and peel results are a property of the core/adhesive/facesheet combination. In the case of failure in the core, the strength of the bond (core to adhesive to facesheet) is greater than the strength of the core, by some unknown amount. Such test results are often reported as "adhesive" strength, but should be reported as "minimum strength of the core/adhesive/facesheet combination"

 

[ESPcomposites]

This has been an interesting and useful technical discussion, but has not really added much to what I already knew. So, I feel I am still left with my original question. Perhaps I have missed it, or maybe I need some clarification.

It is great to know that we CAN predict failure of a perfect bonded joint, provided we have nonlinear adhesive properties as a function of temperature. Researchers (and myself), find it to be an interesting problem.

Stepping aside from the academic world and now wear your engineer's hat. Have we answered the question of why we SHOULD try to solve the problem in this manner. In order to get an accurate result, we need adhesive properties that are expensive to obtain. If an adhesive already has published properties, then we probably already know a good taper ratio/design parameters to ensure adherend failure. Perhaps I am over simplifying?

So there is the chicken and the egg, provided we have a new material system or don't trust the published data. You can either spend money to characterize the adhesive properties, use an analytical model to determine strength, and then test for verification. Or you can spend the money towards directly testing the joint.

Now I did gloss over one issue, which is the proportion of joint strength to adherend strength. For example, if we suddenly changed to base material with a higher/lower strength, then the min taper ratio requirement will also change. But when it comes to CFRP at least, I don't know that there is enough difference to warrant the use of an analytical model.

So perhaps someone can directly point to the value of the analysis, from the engineer's perspective (disregarding the academic value).



Brian

http://www.espcomposites.com

 

[palomc]

@blackmax: Is there a typo in the TN number you gave? Instead of DOT/FAA/AR TN-06/07 should it be TN-06/57 "Best Practice in Adhesive-Bonded Structures and Repairs"?

Some observations. Not to split hairs, but Hart-Smith's philosophy of ensuring that the bond itself is not critical in a bonded assembly is not quite the same as saying that the strength can't be predicted. In order to make the statement that the bond isn't critical (based on analysis during the sizing phase of design), one has to predict it's strength.

In fact, Hart-Smith's methods were used in the late 70's time frame to design the carbon skin/titanium wing root step-lap joint in a well known military product. It wasn't the certification tool, but it was part of the optimization process.

I think the reason why people get away with using average shear allowables from coupon tests in their joint designs isn't completely due to conservatism, it is also partly because the coupon reflects to a certain degree the reality of the repair joint. That is, the coupon test *did* have a peaky shear stress profile, and peel, which can be similar to the actual joint the allowable is used for if the geometry and layups aren't too different. Obviously this is "lucky" and not by design but it's also what happens.

I don't quite get the statement about scarf-joint creep. This may be true for a room-temp adhesive system but I'm accustomed to requiring all the materials in a structure to have a Tg well in excess of the service temp. Now, Tg is not a magic number but still if you remain a good 50F below Tg I wonder if much creep will happen.

Also on that topic, you can think of a scarf joint as a limit condition step-lap joint, with an infinite number of steps. So why would the average shear in a scarf have to be less than the average in a step-lap? I would think, due to less peakiness in a scarf, the average could be pushed higher than a step lap.