Tourist submersible visiting the Titanic is missing Part 2 69

[Tomfh]

The green and blue curves are similar and are superior to the red curves.

The right hand side of the curve is not the "yield" point. They're all failed by that stage. Hence the convergence.

 

[TugboatEng]

I don't know they all still appear to be at 93% of the ultimate strength.

 

[blakmax]

Having read the attached link, I make the point that in both the bolt hole specimen and the ILS specimen in the reference work, there is no possibility of a glossy surface such as would occur in production defects such as probably were referenced in the previous description of the post production defects. The bolt hole and ILS specimens the surfaces that would have been injected would have been induced by fracture of the resin system. Now, provided that the injection process was undertaken soon after the fracture occurred, there is a chance that the exposed surfaces would still be chemically active, so there is a chance that the surface would form a reasonably effective bond.

I would express an opinion that some caution must be taken in extrapolating those results to suggest that the repair method would still produce adequate results if the time after delamination was excessive, or if service conditions enable the possibility of contamination.

It is a very long bow to draw to suggest these results could justify injection of production voids.

Regards

Max

 

[JStephen]

I was googling earlier to see if there were any actual factual updates on this (didn't see any), but did run across a Wikopedia article. Of interest in that article were a BUNCH of reference links, if anyone needs a few weeks' worth of reading.

 

[SwinnyGG]

Do we typically design around ultimate or yield strength?

Is this a serious question? Designing composite structures is not like designing with metals at all.

 

[IRstuff]

The "references" are almost all just news articles; the Wikipedia entry itself is almost shorter than its listing of "references"

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[TugboatEng]

Welcome to the modern age where the factuality of an article is based on how many outlets carry it, even if the story is verbatim from each outlet and they all fire the same sources.

SwinnyGG, I asked a question.

 

[blakmax]

TugboatEng

SwinnyGG's comments about yield/ultimate are correct, if a little blunt (a bit like my responses at times ). In general, composites are considered to be linear elastic to failure if loaded on the fibre direction, but there may be some features that resemble yielding but that is mainly due to resin deformation. The best design methodology is to design to a safety factor below linear limit. All of this changes with layup configurations; more fibres in the load direction or more resin dominated cross plies in the load direction; more shear loading or more axial loading.

But it is not as simple as that. As already discussed, the failure criteria for composite materials is Pandora's Box. The main issue is that failure can be by fibre fracture, fibre buckling, fibre shear as well as resin failure in tension, shear and compression. Then add in fibre disbonding from the resin.

Now I have criticised the Tsai-Hill and Tsai Wu criteria. The principle issue here is that this approach has evolved from the valid Hill failure criteria for ductile metals. This approach for composites is based on the STRESSES in each lamina. It is important to understand that the STRAINS between lamina must be consistent, but because of the different elastic moduli in the fibre direction and perpendicular to the fibre direction, the STRESSES are not consistent between layers. What is consistent between layers are STRAINS.


The issue with the Tsai-Hill and Tsai-Wu failure criteria is that the Hill failure criterion assumes a consistent failure mode (shear) irrespective of the loading regime. The principle criticism of Tsai-Hill and Tsai-Wu criteria is that the failure mode is NOT consistent for all loading regimes. In some cases, failure is by fibre fracture or fibre compression or buckling failure, or even shear failure. As with my discussion about stretching some successful injection repairs to boldly claim that ALL injection repairs are effective, then it is equally invalid to stretch the mathematical model from clear failure conditions to apply the mathematics to universal conditions. As I already pointed out the Tsai-Hill and Tsai-Wu criteria actually predicts a very significant INCREASE in STRENGTH under compression-compression strength as a result of changes in transverse TENSILE strength. I do not understand how a change in tensile allowables can generate an increase in compression strength. It is pure mathematical gymnastics, adn as I said, I hope it did not cause the loss of these lives.

There are strain based criteria (Hart-Smith advocated these).

The real trick with composite failure criteria is to match the failure mode with real test data, and I really acknowledge the excellent work that Mike Hinton et al did with ICCM to expose the reality in the deficiencies in this important aspect of composite technology. My understanding is that work by Jon Gosse (Boeing) actually provides a reasonable transition in failure modes under differing load regimes. But I have left that field of study years ago, so I urge readers to do their own research.

Hope this helps.

Max

 

[SwinnyGG]

SwinnyGG, I asked a question.

It is frequently not clear if you actually want information, or if you're trolilng.

So, ok. The answer at the 50,000 foot level is pretty easy. Generally, for composite structures, if the component has yielded it has failed. The plastic regime of the stress-strain curved is not used, in the sense that you factor ultimate design loads and in service load limits to stay completely out of the plastic regime in all but the most extreme outlier design cases.

So when you're evaluating that chart, you don't particularly care what happens after the peak of the curve, because when you use that criteria for design, you're staying completely to the left of the yield point on the curve. The peak value is what is important.

When you get into damage tolerance/load limits with various damage types/fatigue criteria it becomes very complicated and nuanced, but that's the executive summary.

 

[SWComposites]

Ok, to add to the above 2 excellent posts, for carbon fiber composites (fiberglass composites can be different):
- under in-plane loading (except as noted below), response is essentially linear to failure, so ultimate strength is used.
- for interlaminar failure modes, these are resin controlled, and often involve crack initiation and then growth, so in most cases the initial subcritical failure point is used (sometimes at limit loads and sometimes at ultimate loads, depending on application); also the maximum load is checked against ultimate loads. And in some rare cases people go down the deep dark path of trying to predict crack growth.
- for fastener bearing loads, the response is like shown in the plot above, assuming the critical failure mode is “bearing” and not shear out, cleavage or net tension which are linear to failure. For bearing failure mode, while the response shows “yielding” it is really a local crushing failure. For bearing strength a number of approaches are used. Maximum load is used at ultimate load. 2% Offset load is used at limit load as a quasi yield check. Onset of nonlinearity is used at fatigue load levels to (hopefully) ensure no bearing hole elongation under fatigue.
- there are a few other failure mode types that occur occasionally.
- and then once damage is introduced, residual strength analysis is complicated.

 

[JStephen]

"The 'references' are almost all just news articles..."
Well yes, I didn't say this was a secret trove of unreleased knowledge, in fact, the opposite.
But virtually the entire discussion above and in the previous thread is based on (1) news articles (2) Oceangate's own videos (3) Youtube or magazine interviews with people not directly involved in the project, or (4) Speculation based on (1)(2) and (3). But those references are just putting a lot of this into one list.

 

[moon161]

SWcomposites, I remember detailing a fabricated steel beam and digging thru my omer blodget book and my structures book to find the shear at the web/flange transition, checked it twice and was like, Oh, I'm not going to get a welder to lay down that little metal, am I, and spec'd out 2" in 6 or whatever I had in the rest of the weldment. This becomes completely different if you only have resin transmitting the shear, doesn't it? So would a thick 0° layup have an intraliminar shear failure instead of fibre pullout or fracture when bent?

 

[SWComposites]

moon - possibly, depends on the specific loading, boundary conditions, etc. A short beam loaded in 3 point bending is used to drive an interlaminar shear failure to get that strength property.

 

[moon161]

SWC, Does the radial force gradient cause any trouble with a high pressure vessel like that? You have 5ksi radial at the outer surface and practically nothing at the interior surface.

 

[SWComposites]

"trouble"? hmm, actually, assuming uniform loading (eg, ignoring the end condition stresses for which we don't have enough design details to sort out, but were likely a complicated 3D stress state), then no the pure radial compression stress is not likely a problem. Thru thickness compression strength of carbon/epoxy laminates is ~ 20-25 ksi (tension is much lower).

BUT, this kind of thing would require a very detailed stress analysis of the assembly, and a lot of testing (which Oceangate seems to have not bothered to do). With composites one has to pay attention to stress concentrations, as there is no "yielding" as with metals to save one from design sins. And the apparent presence of fabrication defects complicates things further.

Decades ago I worked a project involving an underwater pressure vessel, for a military application, but the depths were a lot less than the Oceangate submersible, and I (vaguely) recall there were a lot of issues with the end domes and connections to the cylinder sections. However I wouldn't consider myself to have anywhere near sufficient experience to design a composite pressure vessel for the Titanic depths without a lot of fabrication trials and full scale testing; its a completely different loading condition and environment from that which almost all composites people are familiar with.

 

[3DDave]

As far as I can tell, the hull with the noted defects was replaced.

 

[MechinNC]

Not a structural engr here, but I have put a fair amount of thought into this.

Some here seem to have a LOT of experience in structural composites, and I certainly respect that knowledge.

But to this generalist, I came up with a few possibilities:

1. Non uniform distribution of stress in the thick casting in compression. Progressive damage with each dive.

2. Water leakage or adhesive extrusion into edge of cylinder where the end caps engaged. Water or adhesive could extrude into the raw edge of the cylinder.

3. Simple osmosis. Water under that pressure will try (and may) ooze through the surface of the cylinder, getting between windings and weakening them.

4 Portal window blew in.

5. No apparent NDT, either pre or post dive, nothing to quantify progressive damage.




But the biggest gripe I have as a non-expert is the lack of testing. You can build something and not fully understand it. I have done so, quite a few times. But I knew what I knew, and I knew there were things I did not know.

What I did was test. Things failed that I sorta expected (cost factors), things I thought were solid failed or wore. Things I thought were iffy held up like champs. Designs were tuned based on the test results.

But I tested the prototypes and tested them hard.

OG did not seem to have any sort of rigorous test protocol. If there is a big enough hyperbaric water tank test rig, they could have put it there and cycled the pressure. Hundreds of times relatively quickly. Do some NDT after so many "dives" (pressure cycles).

If no tank available, the hire a ship and go north of Puerto Rico and repeatedly dunk the thing. 20k ft depth there. Again, NDT every so many cycles.

Last dive, lower to calculated crush depth and see if your modeling is close.

 

[IRstuff]

OG did not seem to have any sort of rigorous test protocol.

THAT wasn't the issue; Rush didn't want a testing protocol at all, since that would effectively affirm that the naysayers were right and that testing was required to certify safety.

Whatever engineering sense Rush might have, or could have, had was completely subsumed by his, now obviously wrong, ideology that the certification process "stifled innovation."

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[waross]

My feeling is that when innovation runs up against something like a "certification process", there is generally a good reason.
There are exceptions, but it takes a lot of testing to prove the exceptions.
Advice to the younger engineers:
If you think that the codes or testing protocols are wrong, it may it is often an indication that;
"You don't know what you don't know."
I think it was not knowing what he didn't know that did in Rush.

--------------------
Ohm's law
Not just a good idea;
It's the LAW!

 

[JStephen]

"You can build something and not fully understand it..."

To add to this, you can have all sorts of designs based on over-simplified modeling, theoretical modeling without testing, etc.
But there are a couple of issues.
One is, "If in doubt, make it stout", and you can intentionally overdesign to partly allow for potential inaccuracies in the design process. The problem is, when you're trying to skim every last ounce off so the thing will float, this idea sort of defeats the purpose.
One is, consider the consequences. If it costs $200 to properly analyze a part, but only $50 to make it, and the only consequence of failure is to have to make a new one, then yes, take the shortcut and you're likely good. But if the consequence is "Instant gruesome death for all concerned", by golly, that's not the time for any shortcuts, either.