Does anyone know why parts get weaker when you have more of them?

[allmhuran]

I've been doing a little bit of experimenting with "small" craft, where "small" is a relative term meaning "sub component of one of my usual craft".

For example, I've always thought orange tanks had bad (weak) connections. So for the first time ever, I just tried launching a rocket composed of a probe, then 4 stacked orange tanks, then a mainsail (mainsail attached via cubic strut to prevent overheat).

I added zero struts to reinforce the ship.

Now, my past experience has told me that if you stack more than 2 orange tanks, something will break. 99.9% of my failures are "structural failure on link between jumbo 64 and jumbo 64". So I fully expected this rocket to fail as soon as I launched it. To my surprise, I flew it all the way into orbit with barely a hiccup.

OK, I thought, I'll launch 4 at once. So I stuck 4 short I beams radially onto the probe core, and radially attached 4 of the above "4-tank high with mainsail" stacks to the ends of the ibeams. Now, this rocket did fail eventually, but only because it started spinning. In fact, many seconds after it had failed.... at a time when it was falling and spinning wildly, it looked like this (sorry for the night image...)

I know it's hard to see, but if you look closely you can spot the one 4-tank stack still fuctioning with *no breaks* even though the rocket is spinning and flailing around like crazy at this point. And the connection between the "top" orange tank and the girder seems completely indestructible even after twisting completely out of position.

But you'll also notice that there's only 1 and a half stacks there... apparently when there were more parts on the ship, the i-beam attachments were.... less indestructible. And, oddly, the ones that DID break broke when the ship was doing far less crazy shennanigans, so there should have been less stress on the joints.

TL;DR: It seems like fragility is somehow tied innately to part count regardless of structure.

Edited October 12, 2013 by allmhuran

[allmhuran]

Addendum: I have experienced an even more extreme example of this effect, but I doubt many people have seen it because it requires you to build something really, really, REALLY big.

I'm sure most people have at some point attached large solid rocket boosters to the sides of their rockets. I tend to use the "big" radial decoupler under the assumption that it's the strongest. And when people do this, I expect that, for the most part, nothing goes wrong. It's a completely standard thing to build.

But... if you build something really really big. Strange things start to happen.

Take the example below. Now, this is the biggest, heaviest thing I've ever put on the pad by quite some way (there's a million total units of propellant sitting there). But you'll notice that up at the top, around the outsides, are a bunch of large SRB's. They're attached quite typically, just as described above. Nothing different at all.

And yet, on occasion, an SRB just... falls off. ON THE LAUNCHPAD. Before anything has even been turned on. In the VAB you hit the "go to launch" button. You wait for physics to kick in, give things a second or two to stabilize and all of a sudden... wups... an SRB just casually slides off as if, somehow, it knows it's attached to something heavy and, being a lazy SRB, it doesn't want to have to do any heavy lifting.

WEIRD I tells ya.

Edited October 12, 2013 by allmhuran

[Dispatcher]

I see some structures hanging over the launch pad, which do not seem to be supported by it. Do any of them shift or "shake" before you loose a booster while the ship sits on the pad?

[allmhuran]

Not in any particularly extreme way. There are well over 100 launch stability enhancers on the picture in post 2 (many supporting from the inside), so it has, relatively, about the same amount of support on the pad as, say, the 4-tank-plus-mainsail stack that went all the way to orbit with no failures (for which I used 4 launch clamps all attached to the lowest orange tank).

[Dispatcher]

I've had perfectly symmetrical "normally" sized ships with 3 or more LSE's supporting them, which once the physics kick in (on the pad) will shake the ships in an instant. And this is with the connection angles flush with the face of the part being held. I never had that situation until after I updated to 0.21. I'm wondering if this is a related situation?

[allmhuran]

I have noticed that on rare occasions I can actually improve pad stability and, based on anecdotal hunch, structural integrity after launch by *reducing* the number of launch clamps. Best guess, the clamps are so rigid that they don't allow enough flexing to occur when physics is first applied. But I haven't been able to find anything conclusive through empirical means, just hunches.

[Dispatcher]

To clarify: I tend to use a minimum of 3 and a maximum of 8 LSEs.

[SRV Ron]

Those SRB need bracing. In this example with eight SRB when there was no bracing, boosters would twist off in flight. A ring of struts would stop that but they still twisted enough to cause spin issues. Two rings and X braces made the design rock stable in flight. Yet, no braces were needed to the inside ring. The couplers were strong enough to transfer the torque to the core so long as there was no twisting forces to deal with.

As you can see, only a single SAS is used for stability.

[allmhuran]

Nah, that's not the reason. If you can launch a small rocket with unbraced SRBs (and you can), then there's certainly no reason why, on a large rocket, unbraced SRBs would simply fall off on the pad.

[ferram4]

What happens if you move the launch clamps so that they're all arranged along that line of pipes on the orange tanks and clip them through each-other so that you can have the same type of two-tiered clamp system you have going on here? The reason I ask is that the way they're currently attached the upper clamps will tend to pull the rocket counter-clockwise while the lower ones will tend to pull it clockwise; over the length of the rocket that might cause enough movement on physics start to torque an SRB off.

How are those giant orange stacks attached to each-other / the central stack(s)? Perhaps that causes enough movement to knock a booster off.

Try bracing them to see if it works; you're right that you don't need to brace them for a small rocket, but with a small rocket the SRB is connected to a much smaller mass that will move less on the pad while with that monstrosity each SRB is connected to a lot more mass held together by a lot more springs.

Something that might be worth trying with your original post is to retry it, but see if the same part breaks each time; there's a bug where parts added with symmetry don't have the same connection strengths (particularly noticeable with airplanes and wings), so perhaps that is the source of the problem.

[allmhuran]

Good call, I believe the "just falls off" problem is a symptom of the effect mentioned in your last paragraph there Ferram. Now that you've mentioned it I recall that, quite some time ago, I was having a similar issue with a different, much smaller ship. I tried many things until eventually I just removed the offending loose piece and put it back on without symmetry, and suddenly everything was peachy.

This explains the second example, but still leaves the original one open. I suppose the same thing could be happening there (in the 4-stack version), but what really confuses me is the fact that one of the columns has actually *not* broken even with all of the twisting and flipping and whipping around, whereas the others broke while the rocket was much more stable, going fairly straight up but with a mild spin causing the bottom ends to swing out (I deliberately did not strut the bottoms together).

[Person012345]

I have issues on a rocket with a radially mounted fuel tank + skipper engine. Despite having struts they sometimes just snap off for no obvious reason. It's not like a large radial decoupler shouldn't be able to handle the force of a skipper or the weight of a single X200-32 fuel tank.

Edited October 13, 2013 by Person012345

[Kasuha]

You're wrong about things falling off on launchpad before there was any action. There was an action. The ship was moved from zero gravity in VAB to surface gravity on the launchpad the moment the physics engine was started. Based on how your ship is built, various connections will give by various amounts and if some of them get just the right amount of impulse, they will break.

I also don't think that with more parts connections get more brittle. They don't get stronger, that's the problem. With more parts you have longer levers and bigger amounts of force you can put on them. That's why they break. You have to build your ship to distribute forces evenly - both for when gravity kicks in and for when you start engines, stage, etc etc.

[allmhuran]

Yes, you seem to be missing the point there.

If I take a solid booster and attach it via decoupler to a small ship, or a big ship, or the side of the bloody VAB itself, it makes no difference to the decoupler: The weight of the SRB stays the same, and that weight is all the decoupler is required to hold up no matter what happens to be on the other side. So it doesn't make much sense that on a small ship it has no trouble holding that weight (plus the additional G multiplication that of course happens during ascent), but on a big ship it falls off on the pad.

The point about a ship "wiggling" on the launch clamps has already been raised. As I already responded, the big ship doesn't wiggle much on the pad, it has a heck of a lot of support. Even if it did wiggle, we're comparing the amount of stress put on the decoupler by a slight wiggle in an SRB on the pad to the amount of stress when, say, the SRB is actually lit and kicking out thrust while attached to something completely stationary like, say, a launch clamp, where it has no problem. So there's that.

In any case, Ferram already raised a point about a symmetry strength bug, which I had forgotten about, and which I expect is the explanation for what I talk about in post 2.

But the questions raised in post 1 still remain.

[Daishi]

"This wasnt in my job description" - Unity

[SRV Ron]

Nah, that's not the reason. If you can launch a small rocket with unbraced SRBs (and you can), then there's certainly no reason why, on a large rocket, unbraced SRBs would simply fall off on the pad.

All the forces seem to add up, depending on the stack and parts, so they wobble all over the place once set on the pad.

This NovaPunch design was like a wet noodle until it was braced like a exoskeleton.

However, this one has no bracing but flies straight with no wobble. It totals 30 meters of tanks stacked on top of each other and flew SSTE.

So, it appears that such wobble is part and bracing dependent. It could also depend where on the assembly tree they were placed.

[rkman]

Isn't just 'more parts = more mass = greater forces due to gimbaling'?

[Epthelyn]

Isn't just 'more parts = more mass = greater forces due to gimbaling'?

Gimballing forces depend entirely on the engines you're using. If you add 200t to a rocket but keep the same number and type of engines you were using on a 100t rocket the forces will be the same, only the rocket is far more likely to go crazy once it starts wobbling because...er...physics. This has no relevance when a rocket falls apart on the launchpad before you even hit space, though. That's just structural failure/kraken.

Anyway,

It's difficult to tell from the picture but are those orange tanks connected to each other in any other way than the standard connection points? If you put a cubic-octagonal strut or one larger structural parts on either side of the join and then link those two parts with struts it seems to considerably strengthen the connection, probably because structural trusses have limited give to them whilst struts have absolutely none.

The problem with the physics in KSP is that, on rockets this size, you'll often find that no flight ends in the same way as another. If the calculations are done in a slightly different order one side of the rocket could fall off before the other but the reverse could happen in the next launch.

I've never built anything this big so I can't really be of more use than that I'm afraid

[allmhuran]

The pictured rocket is build using a slightly more advanced version of the "thrust plate" technique, but the main feature of that technique persists:all columns are attached to structural parts (i beams, cubic struts, etc). Of course, that's not relevant to the SRB problem, but I think Ferram already hit the nail on the head with that one earlier in the thread.

I've been trying to do some Science on this problem - empirical testing using slight modifications to simple constructions. For example, 9 orange tanks stacked end to end horizontally (in the SPH) with two launch clamps at either end of the resulting column is right on the point of failure. Sometimes when you go to the runway it will break apart, sometimes not. Then I tried putting a structural part, like an I-beam or cubic strut - in between each of the tanks to see what changed.

I also changed the way I built it. Sometimes I attached each orange tank one at a time, and then on other tests I built it by copying a subsection - like building 3 tanks and then copying that twice to make the full column.

The results were inconclusive. Out of 20 tests with "one at a time" construction, the structure failed 4 times. Out of 20 tests with "copy existing" construction, the structure failed 10 times. But that's not really enough tests to be able to conclude that "copy construction" is weaker for some reason.

It also seemed like going to the runway from the SPH resulted in more frequent failures than going to the runway by hitting "revert to launch". But again, the sample size is still too small to be conclusive.

[Moar Boosters]

I think the parts build up a sort of resonating wobble when you slap lots of stuff together. Every joint has a bit of give, and when you put loads of them together, the weight of the vehicle doesn't seem to dampen these forces like it would in real life, instead they build up and shake stuff apart.

It can be seen most graphically if you slap together a massive tower of those tiny grey probe fuel tanks and then attach a launch clamp. Instead of breaking, or falling over, or collapsing under it's own weight...... the thing turns into a bouncing snake, defying pretty much every law of physics we know.

The solution to the problem will hopefully be more realistic physics, delivered in one of the patches. Until then you've just gotta strut the hell of of stuff, because struts don't seem to have as much give as regular joints.