How to heavy lift? How to avoid the Kraken?

[bitzoid]

I've been trying to build heavy lifters. My current goal is to be able to lift 600t out of KSOI and I am using two stages as of now. My lower stage currently conists of about 20 Mammoths. The next about 10 Rhinos (I'm still tinkering).

My problem is that the physics engine keeps eating my ships: No matter how much I strut (auto or otherwise), spatially separate components and keep the vessel steady (touching no controls, etc.) at various points during the ascent the picture freezes for a few seconds and then everything just explodes. My previous designs with more intricate staging were even worse.

What's the weird thing, when I reset to launch it happens much sooner. If I reset again it might even explode before even igniting the first stage, just standing on the launchpad. When I boot up the game from scratch, it usually works for a few km. Once I even managed to get out of the atmo.

Is this a normal thing? I find it very frustrating lately that my larger ships are all being destroyed for no apparent reason. I'm running 1.6.0.2395 (LinuxPlayer).

How can I get around that? How does one build more resilient heavy lifters? (resilient wrt the quirks of the game engine)

[steuben]

Pics of what you've got will help. But I've managed to get 1kt to 1Mm by 1Mm.

What design philosophy are you using?

 

[bitzoid]

Ah, it seems that despite my best efforts some of the large ore tanks that simulated my payload were close enough that the vibrations in the craft caused them to collide. I'll try to separate them further. Edit: This didn't work. See below.

As to your question, I used to prefer a sort of onion staging, but am now down to plain vertical staging. With both stages arranged in a star like pattern. I trimmed it down to a very crude and ugly design. The more involved ones were in a habit of going boom. I'm happy for suggestions how to improve it.

On the top is my simulated cargo (40 full ore tanks).

 

I was unable to make a screenshot from the crash, but it happened outside of the atmo. Exactly when the camera decided to change mode, one part of the payload crashed into the rhino-stage tanks.

Edited January 2, 2019 by bitzoid

[steuben]

Try inverting the design. Take the payload and attach it to the bottom. I worked with the design concept when I was developed the Bell series of lifters. A pic of the base design of them can be found in the link.

The only other suggestion I can chuck out is if you are using autostruts set them to grandparent, maybe root for the lifter. When the engine re-calcs autostruts it can cause weird kraken effects.

[HvP]

45 minutes ago, steuben said:

if you are using autostruts set them to grandparent, maybe root for the lifter.

I second this suggestion.

I have found that it's best to never set autostruts to "heaviest part." When the craft stages the game will detach the autostruts from your lower tanks and grab a different part of the ship. This can cause all sorts of asymmetrical bending and breaking if the craft was flexing at the moment the struts switch over - which in your case it probably will be. Setting a few segments to grandparent part seems to be the most reliable for me. And don't get carried away with the autostruts either; one connecting each stage to the next, or connecting just two segments in long stacks usually does the trick.

Although, for particularly large lifters like that you will need to experiment to find the best results.

Edited January 3, 2019 by HvP

[Tonka Crash]

Are any of the ore tanks clipping into each other? I've seen other people report Kraken attacks with clipping ore tanks. You might try the following for a test payload:

 

[Snark]

Moving to Gameplay Questions.

2 hours ago, bitzoid said:

As to your question, I used to prefer a sort of onion staging, but am now down to plain vertical staging. With both stages arranged in a star like pattern. I trimmed it down to a very crude and ugly design. The more involved ones were in a habit of going boom. I'm happy for suggestions how to improve it.

Well, for one thing, the design you've shown there has hideously terribly awful horrible aerodynamics:

• Big flat payload up front
• Lots of stages down below with big flat surfaces facing forwards.

It's going to take a huge beating from aero forces on the way up.  Which means you're wasting scads of fuel and therefore making the ship much bigger than it probably needs to be.

Just as an example of a rocket that's both ginormous and reasonably aerodynamic on the pad, consider this one (pics in spoiler)

Spoiler

 

5800 tons on the launchpad, 1700 tons to LKO.  Can eject 1200 tons from Kerbin's SOI.  You didn't say how much dV you wanted, but if I take this ship to LKO and then give it a 2000 m/s burn (which is enough to go to Jool), it ends up as a bit over 700 tons.

Flies great, no physics problems. 

[Ozymandias_the_Goat]

15 hours ago, Snark said:

Moving to Gameplay Questions.

Well, for one thing, the design you've shown there has hideously terribly awful horrible aerodynamics:

• Big flat payload up front
• Lots of stages down below with big flat surfaces facing forwards.

It's going to take a huge beating from aero forces on the way up.  Which means you're wasting scads of fuel and therefore making the ship much bigger than it probably needs to be.

Just as an example of a rocket that's both ginormous and reasonably aerodynamic on the pad, consider this one (pics in spoiler)

  Reveal hidden contents

 

5800 tons on the launchpad, 1700 tons to LKO.  Can eject 1200 tons from Kerbin's SOI.  You didn't say how much dV you wanted, but if I take this ship to LKO and then give it a 2000 m/s burn (which is enough to go to Jool), it ends up as a bit over 700 tons.

Flies great, no physics problems. 

I couldn’t help noticing that the design above uses making history, in fact it is almost entirely made of 5m parts. It seems as though @bitzoid‘s design uses only “true” stock parts. Now, you can certainly build very big stock rockets, but the part count will be quite a bit higher that with the expansion parts. If you would like 5m parts for free, there are quite a few mods that will suit you nicely, or Tweakscale to make stock parts bigger.

[Snark]

28 minutes ago, Ozymandias_the_Goat said:

I couldn’t help noticing that the design above uses making history, in fact it is almost entirely made of 5m parts. It seems as though @bitzoid‘s design uses only “true” stock parts. Now, you can certainly build very big stock rockets, but the part count will be quite a bit higher that with the expansion parts. If you would like 5m parts for free, there are quite a few mods that will suit you nicely, or Tweakscale to make stock parts bigger.

Sure.  If your biggest available tank is the stock Kerbodyne S3-14400 (which holds 72 tons of LFO), you'd need more parts to lift the same payload than if you've got big ol' 5m tanks to work with.

But the design principle remains the same.  Build aerodynamically.  Specifically:

• To the degree possible, build tall rather than wide.
• Make sure that each one of your vertical stacks is a smooth, unbroken cylinder (or tapering cone) that has a pointy front end.  Whatever you do, don't have any flat surfaces perpendicular to the oncoming airflow-- that will clobber you.
• Higher TWR will help you-- gravity losses are a killer, and really big ships have the square-cube law working in their favor when it comes to aerodynamics (i.e. drag is less of a concern than with smaller ships).  I've found that a TWR of 2 works pretty well-- if you're a lot less than that, you might want to consider "engining up" a bit.

 

[Mad Rocket Scientist]

Do the explosions happen during staging only, or just at random times? Using the F3 menu might help determine what parts are the first to collide or fail.

[Ozymandias_the_Goat]

36 minutes ago, Snark said:

But the design principle remains the same.  Build aerodynamically.  Specifically:

• To the degree possible, build tall rather than wide.
• Make sure that each one of your vertical stacks is a smooth, unbroken cylinder (or tapering cone) that has a pointy front end.  Whatever you do, don't have any flat surfaces perpendicular to the oncoming airflow-- that will clobber you.
• Higher TWR will help you-- gravity losses are a killer, and really big ships have the square-cube law working in their favor when it comes to aerodynamics (i.e. drag is less of a concern than with smaller ships).  I've found that a TWR of 2 works pretty well-- if you're a lot less than that, you might want to consider "engining up" a bit.

 

Certainly I’m not disagreeing with the technical points. However, it is significant that fewer, larger parts will have less lag, although if your computer is good this is likely not an issue, and the making history engines do tend to be quite a bit better than the stock ones, although it does not appear as if you used them.

[Snark]

2 hours ago, Ozymandias_the_Goat said:

the making history engines do tend to be quite a bit better than the stock ones, although it does not appear as if you used them.

Oh, I used the heck out of them.  That monstrosity I just posted is powered entirely by Mastodon engines-- chosen because they have the narrower variant that allows densely clustering them under a 5m stack.

But if someone is using 3.75m stacks, then they'll be using Mammoths-- maybe with some Mainsails or Twin Boars as asparagus boosters-- and that does the job just great.

[bitzoid]

So ... I finally caved and got MHE. I'm trying to use the c1000 tanks now and am re-designing the ship from scratch. However, I'd like to point out the design I posted was an intermediate model, where I was so frustrated with everything blowing up for no apparent reason that I stopped being bothered with aerodynamics, let alone aesthetics.

And the fact that something explodes outside of atmo when the camera changes angle, indicates that the problem wasn't really with aerodynamics -- yes, the ship was horrible but it didn't blow up because of it. It blew up because of micro-vibration, I guess. I blame floating point errors introduced by heavy load on the engine.

I just wonder why the game runs jittery sometimes, is my i5-8250U insufficient for KSP after all?

[Laie]

20 hours ago, Snark said:

Well, for one thing, the design you've shown there has hideously terribly awful horrible aerodynamics:

• Big flat payload up front
• Lots of stages down below with big flat surfaces facing forwards.

Oh boy, on more critique that begins and ends with aerodynamics. And this isn't even Eve! I've launched similar (and worse) contraptions when it struck my fancy. On large projects, topping off every tank can significantly increase part count; leaving them off costs some delta-V, granted, but that's by no means ruinous.

That said, what concerns me more about the aerodynamics of the vessel are the large rectangular wing pieces high up on the first stage. That far up they're probably more of a problem than an aid.

As for the structural problems, which seem to be the main issue here: don't forget the good old struts. Even in the age of autostruts they still serve a purpose. On this craft I'd use them to secure the outer layers of upper & lower stage against each other.

[Snark]

3 hours ago, Laie said:

Oh boy, on more critique that begins and ends with aerodynamics.

Not "begins and ends", merely "begins". 

Certainly it's possible to launch very un-aerodynamic contraptions, and I suspect we've all done that at one time or another.

However, in this particular case:  the OP's problem is that he's got a ship that is prone to undergoing rapid unplanned disassembly due to some sort of physics glitching.  It's hard to know exactly what's causing the problem, without more information.  However... in my experience, ships that are bigger and more full-of-parts and "contraptiony" tend to be at higher risk for this sort of hilarity-- simple ships usually are fine.  So, one potential path towards "find sometihng that doesn't go kablooie" might be to see, "is there some way to make the ship simpler, smaller, fewer parts?"

The OP has stated the problem as wanting to launch a payload of N tons.  So, the suggestion about trying to improve aero isn't unreasonable-- a more streamlined ship might allow him to launch the same payload with fewer parts, thus simplifying it and maybe alleviating whatever the problem turns out to be.

 

Another option-- which may or may not be acceptable to the OP, depending on what his criteria are-- would be don't do it in one launch.  Currently, he's trying to build a gigantic ship that can loft the payload to LKO and send it onwards to its final destination, all in one launch.  An alternative option would be to build a ship that can loft the payload to LKO... and arrive there with empty tanks.  Then refuel it in orbit (either with fuel-tanker launches from KSC, or with mined fuel from the Mun or somewhere), and then send it on its way.

Doing that is either more tedious or more fun, depending on whether you enjoy the process of sending up tankers, docking, and transferring fuel.    But it would have the effect of allowing the launch vehicle to be much smaller on the pad, which, again, might have a shot at eliminating whatever complexity is causing the freak-out.  Just a thought.

3 hours ago, Laie said:

As for the structural problems, which seem to be the main issue here: don't forget the good old struts. Even in the age of autostruts they still serve a purpose. On this craft I'd use them to secure the outer layers of upper & lower stage against each other.

Yep, no argument there-- they can come in very handy for stabilizing structurally neurotic ships, when applied strategically!

[bitzoid]

@Snark in a sense, your suggestion fixed the symptom by me redesigning the ship. But some of the kablooies did happen well AFTER leaving atmo. They couldn't possibly have been related to the aerodynamics of my admittedly horrible contraption.

 

So, my summary of things I have learned from this thread:

• aerodynamics cannot be dismissed even for testflights
• Autostruts should be grandparent, not heaviest/root (I think this is the single most important thing)
• git good or MHE

Edited January 4, 2019 by bitzoid

[Hotel26]

I don''t have anything directly bearing on OP's particular issue, but I did recently design and test Ivan:

for lifting loads heavier than I've done before.

The test load (460t) is top center and the design specifically seeks to corral it (with real struts) between the triangular outer stack of the lifter.

In addition, auto-struts are somewhat a black art.  The rule of thumb I've used (with airplanes, too) is "Heaviest strut for the central core and grandfather for the peripherals."

Where the Kraken is concerned: "experiment, experiment, experiment with the formula..."  and, yeah...  it has nothing to do with aerodynamics[*].  (or physics)

         

[*] for example:

• pumping fuel in a space station causes it to explode...
• just approaching a space station with another craft CAUSES the space station to spontaneously explode!

Edited January 4, 2019 by Hotel26

[Snark]

3 hours ago, bitzoid said:

But some of the kablooies did happen well AFTER leaving atmo. They couldn't possibly have been related to the aerodynamics of my admittedly horrible contraption.

Sure, but I never said they were.

My point was not that "aero issues are directly causing the hilarity" -- but rather, that by having bad aerodynamics, that forces you to build a bigger, more complicated rocket in order to lift the same payload.  Which, in turn, may contribute to the rocket's predilection for rapid unplanned disassembly.  So the suggestion was "make the rocket more aerodynamic, so that you can make it smaller and simpler, and maybe that will fix your problem."

In any case, glad you've gotten it sorted out. 

[Laie]

4 hours ago, Hotel26 said:

I don''t have anything directly bearing on OP's particular issue, but I did recently design and test Ivan:

Won't upvote because not directly bearing on the issue, but it's a beautiful display of some heavy building techniques. @bitzoid: keep that in mind for when MEH still isn't large enough. The day will come.

[Xd the great]

MOAR struts may help.

[bitzoid]

So, related to my original problem (I have a working lifter by now, thanks to all), I present ... the 5m ...

 

BZ-9V "Stegosaurus" Liquid Fuel Engine

Thrust ASL: 8428.572kN
Fuel: 262.215 LF/s, 320.481 OX/s
Mass: 41.125t
Price: 163.587

[Phreakish]

I struggled with this while trying to get a 5kiloton ship to orbit, and wound up having to rebuild my boosters to get something more stable. I'm really happy with the direction this build has gone, though I could refine it for actual lifting duties, the concept though is pretty sound. 

I've been radially attaching S3-3600 tanks, then rotating them so they form large booster-attachment bosses, then using TD-37 decouplers rather than the typical radial decouplers. I've had a LOT fewer random mechanical disassemblies as a result, and my rockets are a lot more rigid. I use some 'grandparent' auto strutting around these 'bosses' and use a little bit of clipping to make it aesthetically pleasing. 

Sepatrons are too small for boosters this big, so I've clipped 'flea' boosters into them to in order to get decent separation and clearance during flight. It also puts on a good show. 

My pictures below show a 2,200 ton core being lifted. The engine blocks utilize 13 vectors each. But the use of round decouplers in this way has made my very large rockets FAR stronger and easier to fly to orbit.

Spoiler

[StrandedonEarth]

On 1/5/2019 at 12:11 PM, bitzoid said:

So, related to my original problem (I have a working lifter by now, thanks to all), I present ... the 5m ...

 

BZ-9V "Stegosaurus" Liquid Fuel Engine

Thrust ASL: 8428.572kN
Fuel: 262.215 LF/s, 320.481 OX/s
Mass: 41.125t
Price: 163.587

Yeah, those 9-Vector clusters are awesomely powerful, but scarily expensive. I launched one Ludicrous-class lifter with seven 9V cores (I forget what mass it threw to orbit, I think it was in the neighborhood of 800-100t IIRC). Along with the PB-NUKS on the payloads, it came out to over 2M

[Phreakish]

I've done something similar with 17x vectors. 309k credits though! 

I've also been playing with clipping to cluster Rhinos (I hope these get revamped soon and some bare and truss variants added!). The Rhinos ISP makes for a great engine cluster above ~30km, and their thrust is worth 2x a vector at only a 40% cost increase. 
One of my next projects is going to be trying to make the engine blocks recoverable - trying to recover the entire booster is tough without much larger landing legs, but the engines are the majority cost anyway. With a steep gravity turn that puts me around 20degrees over the horizon at 25km I can get my PE to about 20km before even exiting the atmosphere (and I use a lot less fuel to achieve a circular orbit, the drag at 20+km just isn't enough to worry about). That should give me enough time to decouple the expensive bits and attempt to recover some of those credits..