Help with massive launcher!

[ScottyDoesKnow]

Edit: Mission accomplished with a completely redesigned ship! Thanks for all the help.

Original Post

So I'm trying to build a launcher that will take 500t to 75km. Mainly just for the hell of it, I don't currently have any payloads that big. I'm getting close (got it up almost into orbit when the second last stage that's worked a million times knocked an engine off), but for some reason I keep getting breaks between orange tanks. I'm using double stacks of orange tanks in a thrust plate configuration, here's an image to show it better:

I know, it looks crazy at this point, but I can revert a bit if I figure out what the problem is. It's always the double stacks in the row that fail (the 3 forming a T shape on the outside are fine). Originally the double stack was connected with:

3 struts "stitching" between the top and bottom tanks

2 struts at the top to the next inside tank

2 struts at the top to the next outside tank

2 struts at the bottom to the next inside tank

2 struts at the bottom to the next outside tank

1 strut out each side at the top to the next symmetrical tank (or a more inside one if it's too far away)

1 strut out each side at the bottom to the next symmetrical tank (or a more inside one if it's too far away)

Then I started getting the failures, so I added the crazy girder structure in the middle there (two criss-crossed connections to each inside, outside, and symmetrical tanks). I figured that was massively more than enough to stop the top and bottom orange tanks from disconnecting. I also added 4 more struts between the top and bottom tanks for the ones that seemed to fail the most. Still I get 2-5 failures before the first stage is done.

I realize that the ship is ridiculously massive (over 2000 parts), but it seems insane that it fails that much with that much reinforcement. I figure there has to be something I'm missing. Any ideas?

Edited November 22, 2013 by ScottyDoesKnow

[KvickFlygarn87]

Well good day, SDK!

First of all, it would be awesome to see a pic with CoT and CoM enabled.

Second of all, is it the Scotty that made the SDK gun mod for Minecraft? If so, just wanted to say, I love the mod.

[Kasuha]

In my opinion, you should not try to make the structure perfectly rigid. Let it flex a bit instead and strut it so it does some useful work while at it.

Check the design below - the clump of tanks at the bottom was all what was needed to lift 100 tons to orbit. Notice the strutting is pretty basic, what's most important is the single strut going from each tank towards the payload. When the structure flexes, it lifts the payload up. The whole thing is a giant asparagus, tanks are dropped by two.

[allmhuran]

There are cases where adding more struts actually starts to degrade the structural integrity. This is because struts seem to transmit forces and you end up with more forces being transmitted to the weak points of the structure.

What kind of failure are you getting in the double stacks? Is it "structural failure on connection between rockomax jumbo and rockomax jumbo", or is it "structural failure between jumbo and decoupler?".

If the former, then the most likely cause is that the bottom jumbo is wobbling too much, exceeding the elastic limit of the joint. I would remove some struts from other areas and instead strut the bottom of the columns together in a cross stitch pattern. I often do this around the joint as well. Something like this:

The second thing I would do is add the command "disable thrust vectoring" for the mainsails to an action group (at least for most of them) and hit that action group button prior to launch. This will help reduce lateral, "wobbly" forces in the columns.

[ScottyDoesKnow]

Well good day, SDK!

First of all, it would be awesome to see a pic with CoT and CoM enabled.

Second of all, is it the Scotty that made the SDK gun mod for Minecraft? If so, just wanted to say, I love the mod.

Here's the screenshot, but it's all centered so there should be no issue there: http://i.imgur.com/ABpI3wv.jpg And ya that's me, thanks for the love.

In my opinion, you should not try to make the structure perfectly rigid. Let it flex a bit instead and strut it so it does some useful work while at it.

I think you're right about that, I wasn't thinking about what was actually happening. I'm pretty sure I figured it out though, I'll expand more below.

There are cases where adding more struts actually starts to degrade the structural integrity. This is because struts seem to transmit forces and you end up with more forces being transmitted to the weak points of the structure.

So basically I removed all that garbage you see in the middle there. The beam/plate thrust plate is still cross stitched the same but I removed the top angled bar, as well as all the cross stitching between the symmetric tanks. What I found is that now the failure always happens at one of the 6 tanks attached to the center tank, and it's a failure between jumbo and jumbo.

It looks like all the thrust on the outside means it's pushing inwards at the top of the top tank, and pulling outwards at the bottom of the bottom tank. This is shearing the two tanks apart. My basic idea is to somehow connect the bottoms of the outside tanks to the bottom of the center tank, but I can't think of a good way to do it.

And I do have all the thrust vectoring disabled except for the middle 7 engines.

[allmhuran]

I'm having difficulty understanding your description of the new failure. Can you put the craft file up somewhere? I won't be able to check it out soon (currently at work), but in about 5 or 6 hours I can take a look.

In the meantime, from what I can gather you do seem to have a good diagnosis there. The thrust plate helps a lot but it doesn't fix everything, because it does flex a little. It is therefore still better to try and put more thrust under the heavier parts of the rocket, and less thrust under the lighter parts. Since the central column is always the heaviest this means decreasing thrust to weight as you move further radially out.

One way to accomplish this is to add "drop tanks" (with no engines, or only small engines) to the outside ring of the plate. This keeps the edges "weighed down" and prevents them from flexing as much. It's also very efficient because it means your TWR keeps increasing as you stage. A common problem for asparagus designs is that TWR fluctuates between increasing and decreasing as columns get light and then engines get dropped.

What you have at the moment is the inverse of this: a lot of additional columns on the outermost ring providing more thrust there than in the middle. You could move those further in and instead of attaching them as a T-joint, attach them more like:

  C
  |
 /|\
o | o
  o
  |
  o

(C = core)

This would centralize your thrust more and reduce flex in the plate.

It also looks like some of your payload columns have no thrust directly below them. It can of course be hard to get thrust columns there because of space and safe ejection issues, but it does mean that these payload columns that are not supported directly by thrust will be introducing a lot of shear.

Edited November 21, 2013 by allmhuran

[700NitroXpress]

You should never under any circumstances, stack orange tanks on top of each other like that. They tend to not stay together very well, and you're carrying a huge amount of excess weight with you for longer than you need to. Use drop tank designs instead and you can easily get 500 tons into orbit. I've lifted over 700 tons into orbit with that method. Use grey tanks instead and focus all of the energy on the payload.

[allmhuran]

Yeah, you can stack orange tanks (that's what I did for the 750 ton launch) and it can be useful to do so when you would otherwise have space or part count issues, but it can definitely be troublesome. If you do it you have to be really careful about the connection between them, and minimizing any forces that wobble the bottom of the columns.

[ScottyDoesKnow]

Thanks for all the advice guys! Especially allmhuran, I never got a chance to send you the craft file but I appreciate the offer. I tried a couple designs that failed for various reasons and learned a lot. I stuck with the thrust plate design, and moved to single orange tanks.

I tried one with 4x symmetry that just kept wrapping around. Amazingly compact and modular, but due to the wrapping it always spun like crazy.

I then tried 4x symmetry straight out, then added more tanks to the sides of the 4 lines wherever there was room. Unfortunately it was impossible to route the fuel since KSP doesn't like going 2 tanks -> 1 tank -> 2 tanks. I also had a big problem with the sepratrons not working well enough and pieces getting exploded.

Finally I did the same thing but left out the center tank further out so I could route the fuel evenly. I also redid the sepratrons. It works beautifully except that it's 1750 parts and I have to launch at 1 FPS. And it seems that mechjeb has real trouble controlling ships with that kind of lag. When I take manual control though, it actually controls really well when the engines are on and is very sturdy. The stages separate perfectly even when spinning out of control wildly due to mechjeb

Without further adieu, the finished product:

As you can see, there aren't even that many struts holding the payload in place. If you look closely at the visible fuel lines you can kinda see the path they take, basically from the far corner down the side until the middle and then back out to the corner of the next row. The nice thing about having room for so many engines was that I could start with all skippers, open up mechjeb and tweak TWR by replacing them with mainsails wherever needed. The stability would probably allow for another layer, but there's no way my computer will

If anybody wants a copy of the file, let me know. I'll probably be putting it up in the exchange soon anyways though.

Again, thanks for all the help!

Edit: oh ya, one last thing. A huge number of the parts is that each thrust plate panel has 4 struts to the top of the tank. I may tried reducing or eliminating those, but for now I've had enough testing of incredibly laggy ships.

[allmhuran]

WOW. I'm kinda amazed that works, lol. Nice job.

[Fenris]

I think my computer would just melt if I tried to launch that Good Job on getting it to work!

[ScottyDoesKnow]

So I didn't have the little gray tanks on the first launch, it was only about 468t. Just tested it with something real, an unnecessarily massive interplanetary tug/refueling station weighing in at 512.7t. 500 delta V to spare!

Very happy, and very exhausted now.