>>Would you expect carbon fiber to actually do anything useful in compression?
I'd certainly be VERY skeptical of a compression design until it was THOROUGHLY TESTED not only on the bench but in real-world scenarios.
In fact, fiberglass often has better compressive strength and we sometimes add fiberglass outer layers for that purpose, where there is local compressive load.
>>most applications, where the carbon fibers are in tension.
Exactly!
Carbon fibers are astonishing in tension. They take a point load and spread it out through large distances of matrix. Aligning it with tension or bending loads we can make amazing lightweight things.
The matrix is indeed the weak point and the and the shear between layers and between strands is the key failure mode I'm always trying to overcome or evade. Check out this 2min vid on the ASTM Open Hole Compression test [0]; it'll tell you a lot.
I'm not saying that carbon fiber composites are weak in compression, they are actually quite good. But the compression area is usually the point of failure.
The real failure mechanism goes back to the matrix failing between the fibers or layers. A great example was from one product we made using cored construction, an oblate carbon fiber tube with a structural foam core. The goal was to be substantially lighter than aluminum. The standard for that bar was 700 Lbs of load. Our structure was half the weight of the equivalent weight of aluminum, and didn't crack until 1200 Lbs of load. But what cracked was the top layer, in compression. The cool thing was that the rest of the structure held together at much higher loads, because basically the crushed upper load bearing surface was now suspended by the lower carbon fiber in tension, like a hammock.
There is no such opportunity with that sub — the core is the air and the gooey humans inside.
More importantly, the kinds of failure that happen in excess flexion and compression are cumulative.
A key advantage but also very tricky point is that carbon fiber composites do not have the kind of fatigue that metal does. A homogeneous metal still has grain structure, so if it is repeatedly flexed and never approaches the yield point, such as on an airplane wing, micro-cracks will grow and it will eventually fail (they know how to do inspections for this).
Carbon fiber composites with long-chain polymer matrices can flex nearly forever, as long as it never exceeds the yield point at any microscopic point. However, exceed that point, the matrix gets a micro-crack. Those micro-cracks WILL add up, and will accelerate because the entire trick of composites is for the reinforcement carbon fibers to spread the load through the matrix. That failed point of the matrix now no longer contributes to the load-bearing. So, all the other points take the load, and another will fail. This process will accelerate.
The thing is, these failure points can be detected by ultrasound or X-ray examination, but typically not visually, whereas metal fatigue shows up on the surface, especially with the right kinds of paint and/or dyes, so inspection can be easy.
Perhaps most fatally in this catastrophe, is that failure in carbon fiber composites is not gradual or ductile. Metals will show signs of weakness a good amount of time before the failure, and when it fails will tend to bend, not fracture. Carbon composites will fail almost explosively and catastrophically; again, see the ASTM video, the failure at 1:48 is totally typical.
I am actually surprised to read that James Cameron heard from his sources that the Titan crew had dumped their weights and were ascending, trying to manage an emergency. If that was the sensors signalling cracks, I'd expect only milliseconds between the first micro-crack and the catastrophic failure. It seems they had at least seconds, possibly due to the unusually large 5" thick layup.
I could keep going, but yes, you are on the right track.
[0] https://www.youtube.com/watch?v=u1gHxCt-qv0