Comments by "Les" (@les8489) on "Solar Eclipse Timer"
channel.
-
Oh well...let me add my $0.02 here.
First - some observations:
0:19 there are TWO circumferential plies per ONE longitudinal ply. This is as it should be, as the circumferential load is 2x larger. But where are the =/-45 plies ? The rule of thumb is that a laminate should have at least 15% of =/-45 deg. plies. The reason? Shear strength in-plane. Yes, in this case the longitudinal and circumferential stresses may be actually equal, but this is still a good rule to follow.
0:23 "...woefully inadequate glue layers..." Well, sometimes it is better to not say anything than make an incorrect statement. There is nothing wrong with the adhesive...other than a LOT of porosity. Which is to be expected - there is no way to vent the volatiles through 1.0" thick laminate. There may be some peel ply imprints (assuming that the dark areas indicated in the image are resin). But this would not compromise the adhesive, nor diminish its strength - unlike porosity and voids. Square area of the glue has nothing to do with anything - but under bending, inter-laminar shear stress is generated, and in that case the integrity of the glue is critical, as it ensures that all 5 layers (1 to 5) work as a monolithic structure, instead of 5 separate layers.
4:47 The issue of porosity...well, the porosity itself had most likely pretty insignificant impact. The specified 0.46% porosity target is pretty ambitious, and achievable only for thin laminates. The porosity itself (up to about 2%) has practically no impact on strength, and at 4% the deterioration of strength does not exceed 10% reduction (which should be sufficiently covered by applied safety margins). The real problem is that the porosity on a thick laminate is very hard to measure using methods like Through-Transmission or Pulse-Echo. The attenuation for 2% porosity in 1.0" thick laminate is about 40 dB, so for 5.0" it would be about 200 dB. Good luck with that. In short: since measuring the porosity in thick laminate is very hard (if not impossible), the fabrication process itself should be carefully developed and monitored with utmost precision and care. Which hasn't been done, judging by further info in this video.
5:22 I am not sure what that means - but each 1.0” thick layer consisted of about 130 plies or so. I am wondering how many debulking steps were there. I wouldn’t try to compact more than 3-4 plies in one step. This means that each individual 1.0” this layer would have to be debulked at least 30-40 times. Maybe this was done, maybe not, can’t say. But assuming that 40 debulking steps were used, each about 20 minutes, this would translate into 15-20 hours of debulking in total.
5:50 Little porosity in the base plies of each layer, and more at the top layers. Probably not surprising for 2 reasons: ONE - the plies at the base had MORE debulking cycles than the plies at the top of the layer, and TWO: volatiles in the 1.0” uncured layer will migrate toward the top of the layer, as this is where vacuum is applied.
6:37 The multi-cure technique used on the full-size hull…No design manual I know (and I know several) would allow more than 3 cure cycles on the same part (and that on relatively thin laminates). If you need 4th one - then the part lands on a scrap heap. This hull underwent FIVE cure cycles…The reason: thermal expansion factor of fibres is much different than the cured resin. Multiple cure cycles tend to produce microcracks. In addition - I know a process of curing couple of hundred plies which took YEARS to develop. The process of curing epoxy resins is highly exothermic, and the amount of heat produced may lead to carbonization of the resin due to overheating (autoclave temperature limited to 350 deg.F has nothing to do with it, this is due to the low thermal conductivity of laminate and the amount of heat produced during the chemical reaction during the curing process).
7:20 “…no information about the glue…” I would expect that some sort of film adhesive with scrim cloth would be used. From the images, the thickness of the adhesive appears to be quite uniform, so probably must had been the case. I can’t see how manual application of resin/glue could produce a glue layer of uniform thickness.
8:30 Point in case: the bottom adhesive layer was cured 4x, while the bottom laminate layer was cured 5x…no comment here, all that needed to be said was already said above.
8:41 There is nothing peculiar about the curing procedure: it is pretty much standard for a 350 deg. cure system. There is a snag here, however: this is a perfectly OK cure cycle for UNCURED laminate (although the temperature ramp-up is really applicable for rather thin laminate. The rate is specified to allow gradual heating of all plies. As the laminate grew thicker in 1.0” steps - the bottom layers (each 1.0” thick) most likely were NOT heating up at the same rate. This could lead to TWO problems: ONE - the plies at the bottom of the currently cured layer might not heat up at the same rate as the plies directly under the vacuum bag, and TWO: since the 1.0” thick layers were heating up at different rate - this would most likely produce shear stress in the glue between the 1.0” thick layers. Bad idea overall…the multi-cure technique most likely required development of different cure cycles - possibly specific for each subsequent layer (1 to 5). What works for a 0.25” laminate - probably is not so good for a 1.0” laminate piled up on top of 2,3 or 4 already cured layers.
9:04 Yes, yes and yes. Integrity of the glue layers is critical in order to produce a monolithic shell, instead of producing an onion-like assembly of 5 layers. Five 1.0” thick layers with doubtful quality adhesive between them is NOT the same as a monolithic shell 5” thick.
10:25 Peel ply left indentations - yes, this always happens, that’s why this is not always a good idea to use peel ply. The bottoms of these indents are glassy, and adhesion of glue there is pretty poor - which may reduce the shear strength of the glue layer by 20-50%. Nothing new here, mechanical preparation of the laminate for bonding is preferred for highly loaded bonded joints.
12:12 The area of the adhesive has nothing to do with anything…as I said, sometimes it is better not to say anything.
13:26 Yes, that’s what happens when the adhesive fails. You end up with 5 layers 1” thick and having a combined stiffness of 5, rather than a monolithic shell 5” thick with stiffness of 125…In addition (as shown at 0:22), the adhesive was applied at CIRCUMFERENTIAL layers, which was about the worst location - as the shear stress produced by pressure-induced bending acts in the LONGITUDINAL direction. In fact - I would use a layer of +/-45 deg fabric as the interface on both sides of the adhesive layer (which actually is a recommended practice).
Pretty careless engineering as far as I am concerned. YMMV - as usual.
8
-
2
-
2
-
2
-
This is the 3rd video from you I watched. I am speechless...
Yes, you can do some machining of the bumps on the laminate - but not on a structure which works in compression and may be buckling critical. In cases where machining of the cured laminate is performed - this is only done on the sacrificial plies (not structural ones) and is properly accounted for in the analysis. Removal of up to 12 plies is something "innovative" indeed...
This starts being a comedy (despite the tragic end of all people who took the last ride inside the Titan...)
EDIT 1:
LOL...12:10 The guy who was Director of Engineering admits having no background in mechanical or material engineering. Well...I am sure that art degrees fill the gap nicely, and that had a very good DEI program.
Didn't they know that each carbon ply has a nominal thickness BEFORE CURE, which is somewhat larger than the CURED thickness (the cured thickness was measured on EVERY program I was involved in during my 40+ years of engineering practice)?!
This has little impact on a 2D panel (which may have curvature or not), but on a circular structure (even after proper debulking) any reduction in thickness during cure will result in bumps. Interrupting the continuity of fibers by grinding transfers the load to the adjacent plies - in effect, a large stress concentration and a reduction in buckling strength.
Grinding produces visual improvement - but the plies underneath still have a large degree of waviness...this is getting better and better...
EDIT2: In my comments to another video I mentioned that laminates with no +/-45 plies are a definite NO-NO -precisely for the reasons mentioned above. The BOEING guy said that clearly enough - the design manuals at BOEING specify clearly what is the minimum percentage of cross-plies for the laminate to be acceptable in design.
EDIT 3: 17:40 and onwards...very simplified explanation - but spot on in principle. And the funny thing is - to a large extent it could be analyzed. Analysis of damage tolerance, effects of play waviness/discontinuity of plies is nothing new and is routinely done when the designer and the analyst know what they should be doing (and when the accountants agree to the additional expense - which rarely happens). Not always easy, but doable - at least to the extent of pinpointing potential troubles.
But no testing, no analysis of the ACTUAL shell (with defects) and hey-ho, hey-ho, into the depths we go...R.I.P.
2
-
1
-
1
-
@jm8361 Yes, I have seen other videos. It may be true that the only debulked 5 times in the autoclave - as they had to transport it 60 miles to do that. This certainly was a factor, compacting 20 or 30 plies in one go is bound to produce wrinkles.
TORAY spec for the prepreg is 0.0075" thickness. It matches the hull thickness of 667*0.0075 + 4*0.030 = 5.122" after cure.
But it is impossible to lay up, say, 30 plies or so and obtain 30*0.0075 = 0.225" total before debulking. It will be more than that by 10-20 thou at the very least. And when compacted - the stack will settle a bit by delta 't', producing some excess around the circumference equal to 2*Pi*delta 't' - which transforms into wrinkles.
Normal procedure is compaction after 2-3 plies max even on a flat tool, but on a ROUND mandrel I suspect this problem will always be present to some degree.
First 1" layer was OK with only one minor wrinkle: as expected, because the steel mandrel EXPANDED during cure and stretched the layup.
But after 1st cure this didn't work any more: the first 1" layer was rigid, and it practically does not change dimensions during cure. So the problem was getting worse and worse as the radius of each subsequent layer increased.
The 45 deg. plies: if I tried to do this - I would be kicked out of job. These plies are needed to improve damage tolerance, distribute load away from notches and stress concentrations etc.
Another thing: the 1st layer was cooked 5 times. No part goes on the aircraft after more than 3 cure cycles. That's it. Likewise - the 1st adhesive layer was cooked 4 times. Microcracks and delaminations are only to be expected after so many cure cycles.
Another thing: curing epoxy produces heat (this is exothermic reaction). In a thick laminate this may carbonize the resin, as thermal conductivity is low - or may simply deteriorate the resin which will lead to lower properties and failures.
Another thing: subsequent cure cycles were done on the top of previously cured layers. I doubt that the bottom plies in the stack got up to the proper temperature, as they were insulated by the previously cured layers...So - some plies may have been undercooked, some overcooked, and film adhesive might have been damaged even before the 1st dive.
Another thing: they used standard cure cycle for THIN laminates. TORAY is not saying that in their spec - because they assume that idiots are not allowed to make decisions. Thick laminate may require VERY slow temperature increase in MANY MORE steps, otherwise you get temperature spikes and huge lag between the autoclave temperature and the laminate temperature. Thick laminate = SLOW temperature buildup.
Another thing: they shaved up to 12 plies (the bumps). This means that the load from these 12 plies was DUMPED on the first ply in the next layer. Which means huge stress concentration - possibly fiber damage, resin cracks etc.
I did some numbers on the whole design, FEM and hand analysis: overall it is OK: dome thickness, hull thickness, ring dimensions ON THE AVERAGE seem to produce proper structure, which contracts under pressure at the same rate. Some problem is with thermal expansion factors which produce load on the upper flange (this actually helps, as it alleviates effects derived from unequal ring thickness). The inner flange on the ring appears to be much more loaded - at 4000m depth the stress there can go up to 115 ksi. Outer flange is much less stressed - 50 to 70 ksi (these are just very rough numbers).
This thing didn't fail because the design was wrong: it failed because the manufacturing process was done incorrectly, and some incredibly stupid decisions were made. But even if it was manufactured w/o any defects - it would still have safety margins smaller than required. And - strain gauge date from dive 80 should tell anyone with a bit of brains to stop further dives.
I have been doing this (composites analysis, design and manufacture) for more than 40 years and have never seen something equally idiotic. Trying new things - OK. Throwing out of the window all proper practices and cutting corners because money is short - is just...I have no words here short of some unprintable ones...
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
@Dr.TJ_Eckleburg No, the design wasn't that "shitty" at all, albeit it had pretty low reserve factors based on the assumption that the material (that is laminate) has full strength and is manufactured properly. There are several issues here are which make these assumptions invalid:
-several cure cycles (5 for layer 1, 4 for the first adhesive layer, which most likely produced delaminations and microcracks)
-standard cure cycle used for thick laminate (big lag, thermal spikes due to thickness and low thermal conduction, possibly undercooked adhesive, thermal strains etc. - you can write a book on that alone)
-wrinkles/bumps up to 12 plies grinded down (stress concentrations, no 45 deg plies to alleviate the effects)
-insufficient debulking (about 5 (?) per layer)
-thermal cycling issues
-titanium ring bonding and surface preparation (no gloves, no etching, no control of glue filling and thickness at the butt joint etc. It looks like in the axial direction there were some voids there)
-a certain degree of radial constraint introduced by a titanium ring which has variable thickness, and which loads mostly the inner lip on the ring
-thermal expansion mismatch between the ring and the hull (yes, it is getting quite cold at 4000m deep which affects mostly the inner lip on the ring and yes - I did the numbers)
These are just some issues in regard to manufacturing, I could go on - but what's the point? People will still keep saying "it was just glued together" etc.
This sub was destroyed by poor manufacturing process, cutting corners while making it, disregarding the difficulties while producing VERY THICK laminate (different cure cycles with multi-step temperature ramp-up are used for thick stack-ups), having cavalier approach to things like producing play discontinuities by grinding the bumps, possibly introducing thermal damage to resin and adhesive during multiple cure cycles and exothermic phenomena in thick laminate etc. etc. etc. Adding to that - pressure/thermal cycling of both the titanium ring and the laminate.
I am waiting for the final report. I am pretty sure what I am going to see in it - all of the above, plus pressure cycling fatigue on a hull with delaminated layers resulting in buckling and the collapse.
1