How should rockets flex?

[RocketRockington]

Good poll.  I would note that while real world rockets do flex they don't flex like airline wings.  The flex is unnoticeable to the human eye because of the scale of the flex vs the scale of the rocket.  I chose 'perfectly rigid ' mostly because he implementing a realistic flex system that's unnoticeable to the human eye and perfectly rigid are the same thing functionally, whereas zi think most people choosing 'a little flex' still imagine they'd see it

[Superluminaut]

Suggestion thread on topic. Let the devs know what you want.

 

[Brofessional]

The Proton M crash from 2013 is probably one of the best documented examples of a Kerbal style failure in real life. It flips completely upside down before the interstage begins to visibly buckle.

 

 

[Nuke]

i think id rather see autostrut go away with a more elegant system. i always see it as a kludge to fix a rather long running issue with the original game. connecting similar interfaces (same part diameter/size class/part family) should be a lot stronger than connecting dissimilar parts. id say make it automatically enable ridged attachment and possibly come with a strength multiplier (think of it as using more screws to hold them together). with structural parts, i think they should have autostrut enabled when connecting similar sized structures. dedicated structural parts should also be a lot stronger than other kinds of parts of similar size. elegant rules like this make more sense than just turning on all the strengtheners. 

[Casellina X]

18 hours ago, Brofessional said:

The Proton M crash from 2013 is probably one of the best documented examples of a Kerbal style failure in real life. It flips completely upside down before the interstage begins to visibly buckle.

 

 

I'm not sure why they didn't revert to VAB 

[TheoBrandt]

I don't know much about the irl engineering on the subject. Though I think to get into that kind of misses the point. KSP is clearly a game aimed at a young audience, trying to turn it into a true to life simulator kinda distracts from its roots and the cartoonishly bendy rockets are a key part of the games' character. Having said that, us adults who also fell in love with the game want a more mature simulator experience and there's nothing wrong with that either, why can't it be both? If rigidity is something easily tweaked in configs anyway, couldn't we have a "sturdy rocket" option in campaign creation so that those of us who want more rigid physics can have them?

[TheOtherDave]

Just as a quick note about rockets breaking up when they go sideways… they don’t all break when they go sideways. Starship does its re-entry braking by flying sideways. It is true that none of them materially wobble, though. I would say that if the flexing ever becomes noticeable, the rocket should probably just break up instead.

[RCgothic]

On 3/5/2023 at 11:47 AM, TheoBrandt said:

I don't know much about the irl engineering on the subject. Though I think to get into that kind of misses the point. KSP is clearly a game aimed at a young audience, trying to turn it into a true to life simulator kinda distracts from its roots and the cartoonishly bendy rockets are a key part of the games' character. Having said that, us adults who also fell in love with the game want a more mature simulator experience and there's nothing wrong with that either, why can't it be both? If rigidity is something easily tweaked in configs anyway, couldn't we have a "sturdy rocket" option in campaign creation so that those of us who want more rigid physics can have them?

People keep saying bendy rockets are a "feature" that younger players appreciate, and I just can't get my head around how *anyone* would prefer a feature that effectively disables the stability assist and adds this much frustration to controlling larger rockets.

Who likes building large rockets? Everyone.

Who likes trying to fly a wet noodle that will crash no matter the skill of the pilot? Is there even one person?

Edited March 13, 2023 by RCgothic

[Kerbart]

The problem with realistic flex is that it's so little that you don't see it. Metal is elastic; everything flexes under stress. But the displacement it causes is so little that we rarely see it in daily life. Apply that to rockets and they would just break. Visible flex is a bit like an early warning system that you're asking too much of the construction. The problem with KSP is that parts don't flex, only joints. Not only do joints have to be overly flexible to compensate for the lack of flexing in the parts themselves, we also have a lot of them. A real-life booster is a single mechanical entity and not cobbled together from two or three tanks.

One of the problems is that most people are, truth be told, lousy mechanical engineers. Everybody has their mouth full about the need for more realism with n-body physics, Lagrange points and decaying orbits, when when it comes to unstable contraptions the response is not "let's build something better" but rather "We NeEd AuToStRuTs."

I've never used autostruts. I use regular struts where it makes sense (Medium - Extra Snall - Medium? You bet I connect the medium parts with struts). I don't launch a 5m payload on top of a tall 2.5m rocket. My rockets don't noodle.

I'm not saying the current setup is satisfying—far from that. Yes, there is too much noodling but the correct solution is make the rocket break apart, not "make the joints stiffer" or "we need autostruts." It means that you can't just build anything you want and expect it to make it to orbit. That's not always easy. But that's why it's rocket science.

[Periple]

I agree @Kerbart! The rockets should just RUD if overstressed, the flexing is unnecessary. It would also make the physics a lot simpler and scale better to really big vessels, and banish at least one kind of kraken -- the one that causes harmonic oscillations!

[Kerbart]

7 minutes ago, Periple said:

I agree @Kerbart! The rockets should just RUD if overstressed, the flexing is unnecessary. It would also make the physics a lot simpler and scale better to really big vessels, and banish at least one kind of kraken -- the one that causes harmonic oscillations!

There's also a tendency for players to model within the boundaries of the game, not within the boundaries of reality. Partly that's ok, the game only allows for so much and the atmo model is not accurate. But at the same time if you play a little bit more within "would this work in real life" constraints you avoid surprises later on.

In KSP1 the launch profile often used was "straight up to 10 km, then slam the ship in a 45° pitch" which was most efficient with the pre .95 souposphere. I always applied a gravity turn because that procedure just seemed ugly to me, but the number of complaints on how the game "was broken" since it now it was now "impossible to launch anything into orbit, my rockets spin out of control for no reason..." Same for descent profiles that allowed entering the atmosphere at a 90° angle at 4 km/s—you'd automatically slow down to 100m/s before opening your chutes anyway. Not that anyone cared because you could have your chutes deployed at hypersonic speeds without ill consequences. It's like using a Mk-I lander can for atmospheric reentry. Yes, the game lets you get away with it. That doesn't mean you should do it, though; and don't complain when the game is fixed and doesn't let you get away with it.

Noodling is something similar but it got never addressed the right way. It's not an issue if you launch something that doesn't noodle in the first place. Stop launching noodle rockets, and the problem goes away!

[RCgothic]

"Don't build rockets that look like F9" is a bug, not a feature.

[Kerbart]

1 hour ago, RCgothic said:

"Don't build rockets that look like F9" is a bug, not a feature.

Show me the F9 with a 12m payload on top.

[Periple]

I don't get it, F9-like rockets work just fine! It's not a problem if the fairing is a bit wider than the rest of the rocket if it's somewhat pointy and the CoM stays well ahead of the CoP. Just don't make them look like Megamind!

[dvrabel]

Visibly flexing rockets gives a good visual indicator that the rocket is over-stressed and may be about to fail. Without some sort of indication like this, rockets that spontaneously break apart with no warning will be very frustrating , particularly for less experienced players.

I think KSP2 should:

1. Provide a difficultly option for rigid nodes (alongside the existing option for unbreakable nodes, or combined together), for those that want this style of gameplay.

2. Improve the SAS controller in atmospheric flight (it is currently underdamped which makes it more likely to introduce oscillations).

3. Fix bugs where some parts are more flexible than others in the same size class (i.e., the upcoming engine plate fix).

4. Provide parts (e.g., the truss structures) that are stronger (less flexible) but have other trade-offs (such as worse drag or bad thermal performance).

5. Guide players to constructing large, complex craft in orbit (where rockets are not subject to aerodynamic forces).

6. Consider toning down the amount of flex to ease control issues while still providing a visual indication of high stress and impending failure.

[thewhitemetroid]

The Idea of "Joint rigidity" is silly to me because this is not how it works in real life. the weld seams are stronger than the base metal. If a failure occurs it will happen near the weld or at some other location. So having joint flex occurring where a small and medium sized tank join should not be a thing in this game at all. I get why the devs made "tanks" the way they did in KSP 1, but at this stage they should all be procedural, because as Kerbart stated, this is not how a vessel is constructed. Carbon steel, Stainless, Alloy sheets come in standard sizes with an upper and lower limit, and the number of weld seams are determined by the engineering needs of the project and the size of the sheets, not the size of the "gas tank". I understand pressure vessel design and the design for rockets will use similar considerations, but not entirely since wind shear, drag, and many other considerations must be made. But a cylinder is a cylinder. Speaking of which, I agree with most here, noodley rockets are not realistic. If a real steel cylinder was subjected to the same flex seen in KSP 2 the compressed side would just crumple and subsequently the rocket would start to break apart.

As far as what would happen when a rocket is sideways in atmosphere I can't say because I have no experience there but clearly some have managed it, others have not.

Lastly, I have never had fun with a "spinning rocket". 

[dvrabel]

I'm not sure the "real rockets don't X." argument really works in the context of KSP since considerable efforts goes into real rocket design to ensure they don't fail, whereas KSP allows you to cobble-together any old parts into something vaguely rocket-like. If you did this in real life, you would make any awful lot of explosions.

Lots of KSP rocket designs should fail (this is a key gameplay loop), and thus KSP needs to provide useful, more intuitive feedback to the player about why their rocket failed.

[Periple]

1 minute ago, dvrabel said:

Lots of KSP rocket designs should fail (this is a key gameplay loop), and thus KSP needs to provide useful, more intuitive feedback to the player about why their rocket failed.

I agree!

In my experience though, flexing rarely provides such useful or intuitive feedback. My craft fail most commonly because of aerodynamic problems, and they manifest by the vessel spinning out of control, then RUDding because of aerodynamic forces. The intuitive, useful feedback is the loss of control, not the flexing. 

I'm not claiming that flexing never gives any useful feedback at all, I'm sure there are some instances when it does -- but I do think that these are less frequent and less important cases, and the downsides are more significant: kraken attacks, more expensive and worse-scaling rigid-body physics, and structures that really oughtn't be flexy but are anyway.

[thewhitemetroid]

I agree it would be unrealistic to expect a video game to perfectly mimic real life but that doesn't mean it shouldn't try in a simulation game. 

I feel the failure indicator should be "rocket exploded" because that is the real life implication of bad rocket design or an unknown flaw. And at that point you likely didn't lose much since most of these instances are going to be in atmosphere. We can always do the realistic thing and "revert to VAB". The unfortunate thing at the moment is that it's easy for us to blame it on a kraken attack instead of suboptimal rocket design. If I as a learning user can get a noodly rocket into orbit then there is lost potential to motivate me to seriously reconsider my design. Besides, noodly rockets just suck.

[dvrabel]

45 minutes ago, Periple said:

more expensive and worse-scaling rigid-body physic

I must admit that my structural engineering is seriously rusty (only studied some at university decades ago), but I'm struggling to think of a model for a fully rigid structure that would be computationally less expensive than the current model, yet still calculates stresses throughout the structure.

The only technique that I can recall is somewhat similar to the existing model except instead of actual angular or linear displacements to determine forces (via a simple linear spring model) , it uses small/infintesimal displacements and would thus seem to have similar computational complexity.

But as I said, I forgotten much of my structural engineering knowledge, so If someone is more knowledgable, some pointers to other models would be greatly  appreciated.

[Kerbart]

1 hour ago, thewhitemetroid said:

I agree it would be unrealistic to expect a video game to perfectly mimic real life but that doesn't mean it shouldn't try in a simulation game. 

The way it's done in discrete mechanics is representing a solid body as a collection of point masses connected to each other with springs and dampers. From what I can tell in Unity those point masses are then replaced with parts connected through joints. It's far from a perfect model but within the constraints of "lego style building" and reasonable performance probably a good approximation. And more importantly: it's what we're stuck with.

This is one argument in favor of procedural tanks. In itself having weakish joints is not a bad proxy for the "things gonna flex" design paradigm, but by nature each joint will multiply the flexibility in a system and you're forced to do so; you quickly need multiple T800's or Jumbo's stacked.

Here's an idea for the devs that should be relatively easy to implement: automatically increase joint rigidty to "very high" when connecting identical parts (eg T800 to T800) so they behave as one.

1 hour ago, thewhitemetroid said:

I feel the failure indicator should be "rocket exploded" because that is the real life implication of bad rocket design or an unknown flaw. And at that point you likely didn't lose much since most of these instances are going to be in atmosphere. We can always do the realistic thing and "revert to VAB". The unfortunate thing at the moment is that it's easy for us to blame it on a kraken attack instead of suboptimal rocket design. If I as a learning user can get a noodly rocket into orbit then there is lost potential to motivate me to seriously reconsider my design. Besides, noodly rockets just suck.

Good point. The counter argument is that "boom" is a lot less educational than noodling when it comes to bad rocket design, at least in KSP1. The problem is that right now, as you point out,  everybody yells "It'S ThE KrAkEn" for everything they don't like. For me, Kraken attacks are when floating point errors go berzerk and that's in a very limited amount of cases.

Not that anyone will listen, but what they can do:

• Increase rigidity
• Lower joint tolerance for stresses
• When reaching, say 75% of failure levels (or whatever works) start playing animations of bits and pieces popping off as an indication that things start breaking. Or maybe even at 100% but have some kind of "failure" buffer that starts filling (faster the more you go over max level) and then break once the buffer is full.
• You'll end up with visual clues that things are going to break apart before they actually do

I fully agree with you that failure is a better teacher than noodling. Especially when rocket designers start blaming noodling on the game, not on questionable design.

 

[Kerbart]

1 hour ago, dvrabel said:

I must admit that my structural engineering is seriously rusty (only studied some at university decades ago), but I'm struggling to think of a model for a fully rigid structure that would be computationally less expensive than the current model, yet still calculates stresses throughout the structure.(...)
But as I said, I forgotten much of my structural engineering knowledge, so If someone is more knowledgeable, some pointers to other models would be greatly  appreciated.

You bring up a very good point. In fact,@Nate Simpson should consider that for the tutorials. Mechanical design is as much a part of rocketry as celestial mechanics are. Knowing what increases stiffness is easy, probably even intuitive, for those who are trained at it. Knowing why a tube is nearly as strong as a solid pole, why an H-beam is much stronger in one orientation than in 90° rotated, etc. For me, achieving stiffness is usually a matter of placing a few struts. At the right places.

If you don't have a background like that you can make your ship look like a porcupine with struts and still have it noodle. "Struts don't help, the game sucks." KSP does a very good job at punishing wrong designs. What's lacking at the moment is it's ability to teach that (other than showing the results) and with the amount of bugs in the game it's easy to blame the game for that, instead of one's design.

[darthgently]

8 minutes ago, Kerbart said:

The way it's done in discrete mechanics is representing a solid body as a collection of point masses connected to each other with springs and dampers. From what I can tell in Unity those point masses are then replaced with parts connected through joints. It's far from a perfect model but within the constraints of "lego style building" and reasonable performance probably a good approximation. And more importantly: it's what we're stuck with.

This is one argument in favor of procedural tanks. In itself having weakish joints is not a bad proxy for the "things gonna flex" design paradigm, but by nature each joint will multiply the flexibility in a system and you're forced to do so; you quickly need multiple T800's or Jumbo's stacked.

Here's an idea for the devs that should be relatively easy to implement: automatically increase joint rigidty to "very high" when connecting identical parts (eg T800 to T800) so they behave as one.

Good point. The counter argument is that "boom" is a lot less educational than noodling when it comes to bad rocket design, at least in KSP1. The problem is that right now, as you point out,  everybody yells "It'S ThE KrAkEn" for everything they don't like. For me, Kraken attacks are when floating point errors go berzerk and that's in a very limited amount of cases.

Not that anyone will listen, but what they can do:

• Increase rigidity
• Lower joint tolerance for stresses
• When reaching, say 75% of failure levels (or whatever works) start playing animations of bits and pieces popping off as an indication that things start breaking. Or maybe even at 100% but have some kind of "failure" buffer that starts filling (faster the more you go over max level) and then break once the buffer is full.
• You'll end up with visual clues that things are going to break apart before they actually do

I fully agree with you that failure is a better teacher than noodling. Especially when rocket designers start blaming noodling on the game, not on questionable design.

 

Progressively louder and more alarming bad noises leading up to failure would be great.  Metallic groans, sheet metal popping out of shape, something rattling/flapping in wind, rivets popping, then very loud sound of big metal snapping.  All prior to "boom"

[thewhitemetroid]

23 minutes ago, Kerbart said:

• When reaching, say 75% of failure levels (or whatever works) start playing animations of bits and pieces popping off as an indication that things start breaking. Or maybe even at 100% but have some kind of "failure" buffer that starts filling (faster the more you go over max level) and then break once the buffer is full.
• You'll end up with visual clues that things are going to break apart before they actually do

 

9 minutes ago, darthgently said:

Progressively louder and more alarming bad noises leading up to failure would be great.  Metallic groans, sheet metal popping out of shape, something rattling/flapping in wind, rivets popping, then very loud sound of big metal snapping.  All prior to "boom"

Both great ideas for indicators, and sounds like something that would actually add some fantastic immersion and tension to the launch. I would add that we could also add vibration to the navball possibly as an early indicator since excessive vibrations will likely accompany failure events.

[darthgently]

29 minutes ago, Kerbart said:

For me, achieving stiffness is usually a matter of placing a few struts. At the right places.

Does the physics model take triangulation of manually placed struts into account?  I do it as it makes sense IRL