Starship heat shield

[joek]

Yes, the tiles are  ~-05% fiber and binder and 95% void.

For attachment points, maybe when they are casting/molding the tiles, they insert a fiber matrix that terminates in one or more attachment points?

Or they could glue an attachment plate onto the bottom of each tile, but that could add a lot of weight.

Likely those attachment points are what we are seeing in a couple photos--they appear to go directly from the airframe to the the TPS.  That appears to be one possible difference between previous applications of, e.g., TUFROC on X-37B; to my knowledge we have never seen what the airframe attachment looks like.  However, public presentations show attachment between two layers of the underlying TPS (see previous post), which might be similar to the airframe attachment.

That said, it is unclear if the types of attachments shown are semi-permanent and difficult to easily replace, or require substantive work to replace.

Think that difference is worth noting: How much time-cost-effort-$ does it require to initially fabricate-install , versus how much time-cost-effort-$ does it require to maintain-replace?  Obviously robustness of solution affects cost of both.  Guess is that SpaceX would prefer a more robust solution that decreases maintenance.

Which is why we are seeing what appears to be a base level airframe-to TPS pin-based (pop-on, pop-off) arrangement.  All they need to do is reduce the heat transfer/flux from the lower level TPS to the airframe to an acceptable level.[1]


[1] edit: p.s. Not to minimize the problem; certainly a challenge.  But has to be considered as a whole.  For example, Starship's "fluffiness" should help reduce TPS demands.

[envy887]

In my design the tiles get narrower as you move towards the apex. At some point depending on the degree of curvature relative to the size of the tiles it might also be necessary to decrease their height as well.

It looks to me like the radius in the in that direction is constant, so you would not need to decrease the height. Getting narrower is enough.

[aero]

In my design the tiles get narrower as you move towards the apex. At some point depending on the degree of curvature relative to the size of the tiles it might also be necessary to decrease their height as well.

It looks to me like the radius in the in that direction is constant, so you would not need to decrease the height. Getting narrower is enough.

Are you considering that the heat shield only needs to be on one side? That is, they only go ~halfway around. That means that the ends of a row of tiles need not align with the previous row. Does that mean that the same sized tiles could be used for adjacent tile rows? Maybe not, but it is an additional degree of freedom to consider.

[john smith 19]

I have a questions based on excellent BCG's photos https://forum.nasaspaceflight.com/index.php?topic=48895.msg2072270#msg2072270

There are indications that some tiles on SS SN4 are made by coating bent sheet metal. There are even some areas where there is small surface damage to the coating. Is there any other explanation?

The Shuttle had thin high temperature pads which would be inserted in the gaps between tiles to prevent heat getting under the tile.
Elon is trying to avoid having to do this by using hexagonal tiles, however he may have to resign himself to there being a need for something to be positioned between the tiles.

They were called "gap fillers"  and were put in to stop rentry plasma hitting the aluminum skin.

Then the tiles were glued to thin nylon fabric layers (AIU like the stuff womens tights are made out of) to handle the contraction and expansion.

But steels TCE is substantially less than aluminiums so such a support pad should be redundant.

Also is it my impression that each tile is on a single point mounting? In principle each tile could tip and tile (slightly) in the airstream. This might be enough to avoid flow separation and any need to curve every tile to match the underlying surface.

But this is the easy part.

Wrapping that TPS around the wings/drag surfaces/whatever is likely to prove much trickier.

[DusanC]

I have a questions based on excellent BCG's photos https://forum.nasaspaceflight.com/index.php?topic=48895.msg2072270#msg2072270

There are indications that some tiles on SS SN4 are made by coating bent sheet metal. There are even some areas where there is small surface damage to the coating. Is there any other explanation?

The Shuttle had thin high temperature pads which would be inserted in the gaps between tiles to prevent heat getting under the tile.
Elon is trying to avoid having to do this by using hexagonal tiles, however he may have to resign himself to there being a need for something to be positioned between the tiles.

They were called "gap fillers"  and were put in to stop rentry plasma hitting the aluminum skin.

Then the tiles were glued to thin nylon fabric layers (AIU like the stuff womens tights are made out of) to handle the contraction and expansion.

But steels TCE is substantially less than aluminiums so such a support pad should be redundant.

Also is it my impression that each tile is on a single point mounting? In principle each tile could tip and tile (slightly) in the airstream. This might be enough to avoid flow separation and any need to curve every tile to match the underlying surface.

But this is the easy part.

Wrapping that TPS around the wings/drag surfaces/whatever is likely to prove much trickier.

SS SN04 tiles have 2 types of snap mounting:
1. On all 6 sides
2. On 3 studs

[john smith 19]

Think that difference is worth noting: How much time-cost-effort-$ does it require to initially fabricate-install , versus how much time-cost-effort-$ does it require to maintain-replace?  Obviously robustness of solution affects cost of both.  Guess is that SpaceX would prefer a more robust solution that decreases maintenance.

Indeed.

Shuttle tile replacement costs were listed (on the TPSX database, now defunct) as $12000/m^2

However SS/SH will be radically cheaper as it is clear (unlike with shuttle) there will be a substantial amount of commonality, rather than the Shuttle, where nearly every tile was unique in size, shape and thickness (thinking this was a good idea may have been the design teams biggest failure. Actually building it was possibly the greatest triumph of the manufacturing team).

[Slarty1080]

Having a large number of intricate tile shapes might not be a problem by itself as long as they don't require a manual repairs. Unlike on the Shuttle there is no chance of getting hit by debris during launch.

For tiling a rounded cone they could do it using the same slightly warped hexagon at each height point.

But they might eventually go for unique shapes of varying thickness anyway in order to optimize mass.

It's actually mathematically impossible to tile a rounded cone with any sort of uniform warped hexagon.  If you try it, you'll find that it works alright at the beginning, but as the surface continues curving, your hexagons will become more and more distorted until you're forced to add pentagons to the mix.  This is a direct consequence of the Euler Identity.

There's just no way to avoid requiring a large number of different tile shapes when trying to tile a surface with non-zero curvature.  (A cylinder has, mathematically speaking, zero curvature, just to be clear on this.  Not so for a rounded cone)

Mathematically you're correct.

But engineers like to cheat

1 type of tile in 2 forms, whole and cut in half.

PS: I'd like to emphasize that only 200deg of SS needs to be tiled.

Long straight lines like that are problematic.

In the words of Elon when asked why hexagons:

No straight path for hot gas to accelerate through the gaps

https://twitter.com/elonmusk/status/1107379727302451200

All of the verticals are the same size as the standard hex tile and the other lines are shorter.

[john smith 19]

SS SN04 tiles have 2 types of snap mounting:
1. On all 6 sides
2. On 3 studs

Thanks for that detail.

[joek]

...
Also is it my impression that each tile is on a single point mounting? In principle each tile could tip and tile (slightly) in the airstream. This might be enough to avoid flow separation and any need to curve every tile to match the underlying surface.
...

Not sure I see that... Would guess it depends on required heat load transferred to (or isolated from) tiles to airframe.  If heat load can be limited to tiles (no xfer to airframe), would not matter; if heat load is required to be transferred to airframe, then likely does matter as those attach points will need to xfer heat load.  Then we are all into guessing about what level of heat load can be transferred or can or should be transferred from tiles to airframe through those points.... I have no idea.

[joek]

SS SN04 tiles have 2 types of snap mounting:
1. On all 6 sides
2. On 3 studs

Please provide a definitive reference?  We have conjecture as to attachment for both.  Caution that what we have seen is not necessarily indicative of final configuration.  Thanks.

[DusanC]

SS SN04 tiles have 2 types of snap mounting:
1. On all 6 sides
2. On 3 studs

Please provide a definitive reference?  We have conjecture as to attachment for both.  Caution that what we have seen is not necessarily indicative of final configuration.  Thanks.

Only true reference except EM

BCG

https://forum.nasaspaceflight.com/index.php?topic=48895.msg2070295#msg2070295
2nd and 3rd pic, zoom in.

Also, I said SS SN04 so OFC not final configuration.

[Bananas_on_Mars]

If there are 100 rows, each with indentical tiles, and each tile has the same fastening arrangement, then every tile on the SS can be cut from the same identical raw tile (using a water jet cutter).
It would be to much weight to take 100-200 spare tiles to mars. Instead they will just take half a dozen or so of these raw tiles as spares and use simple hand tools to cut them to any of the 100 or so different shapes (corresponding to 100 different rows).
They would have to get some practice in custom making tiles before they left for Mars.

For „field repairs“ on mars they could easily use dissimilar materials (for example ablative materials). AFAIK Elon also said mars ships might have a heatshield that‘s different from what your average daily-use starship will have. Most Starships flying to mars might only use their heatshield once.

[john smith 19]

BTW

A few years ago DLR ran a sounding rocket test with the Brazilian space agency.

They built a nose cone out of standard sized RCC tiles, mostly triangles, pentagons etc.

The key development was the software to confirm despite the segments being flat that the aerodynamics were OK and the flow was smooth enough to avoid turbulent flow (which multiplies heat transfer between 4x and about 6x).

[Slarty1080]

In my design the tiles get narrower as you move towards the apex. At some point depending on the degree of curvature relative to the size of the tiles it might also be necessary to decrease their height as well.

It looks to me like the radius in the in that direction is constant, so you would not need to decrease the height. Getting narrower is enough.

Are you considering that the heat shield only needs to be on one side? That is, they only go ~halfway around. That means that the ends of a row of tiles need not align with the previous row. Does that mean that the same sized tiles could be used for adjacent tile rows? Maybe not, but it is an additional degree of freedom to consider.

The problem is that each tile needs to align with those in front and behind it. On a curved surface this means the angles on the hexagons have to change (see my diagram at the top of page 1) unless the intention is for tile overlap, in which case regular  hexagons might be tried, but I think it would be very messy. The areas of overlap would change and the zig-zag peaks and troughs between layers of tiles would run out of phase and wouldn't line up.

[wes_wilson]

Is the flow over the tiles uni-directional enough to do overlapping tiles?  If the tile was thickest at the front where it attached and tapered towards the overlapped portion (like a wedge) you could keep the actual surface fairly close to smooth and use a fish dragon scale shape. 

https://www.featurepics.com/FI/Thumb300/20090104/Big-Carp-Scales-1022424.jpg

[john smith 19]

The problem is that each tile needs to align with those in front and behind it. On a curved surface this means the angles on the hexagons have to change (see my diagram at the top of page 1) unless the intention is for tile overlap, in which case regular  hexagons might be tried, but I think it would be very messy. The areas of overlap would change and the zig-zag peaks and troughs between layers of tiles would run out of phase and wouldn't line up.

Practically there are 2 issues here.
1) Finding the minimum number of tile type and sizes that map the surface. Kind of like how you render a CAD image with the minimum number (and type) of polygons.
2) How the edges of those polygons line up.

1) is relatively simple (for simple shapes)
2) Gets much more complex. The real question is how good does the polygonal approximation to the underlying surface need to be?

Because an exact match would imply thick tiles that can be machined to a curve (which is how they did it with the shuttle, and part of the very high cost of maintenance).

My instinct (and the work of DLR sounding rocket test) is that is not necessary provided the discontinuities are a fairly small part of the boundary layer thickness

If so you can get away with flat files of the same shape, but multiple sizes, all machined from a standard size billet.

That problem gets a lot trickier on the fin leading edges, with high curvature, needing a lot of small (but IMHO still standard shaped) tiles to handle the leading edge.  The area between the body and the control surfaces being exceptionally tricky.

[Slarty1080]

The problem is that each tile needs to align with those in front and behind it. On a curved surface this means the angles on the hexagons have to change (see my diagram at the top of page 1) unless the intention is for tile overlap, in which case regular  hexagons might be tried, but I think it would be very messy. The areas of overlap would change and the zig-zag peaks and troughs between layers of tiles would run out of phase and wouldn't line up.

Practically there are 2 issues here.
1) Finding the minimum number of tile type and sizes that map the surface. Kind of like how you render a CAD image with the minimum number (and type) of polygons.
2) How the edges of those polygons line up.

1) is relatively simple (for simple shapes)
2) Gets much more complex. The real question is how good does the polygonal approximation to the underlying surface need to be?

Because an exact match would imply thick tiles that can be machined to a curve (which is how they did it with the shuttle, and part of the very high cost of maintenance).

My instinct (and the work of DLR sounding rocket test) is that is not necessary provided the discontinuities are a fairly small part of the boundary layer thickness

If so you can get away with flat files of the same shape, but multiple sizes, all machined from a standard size billet.

That problem gets a lt trickier on the fin leading edges, with high curvature, needing a lot of small (but IMHO still standard shaped) tiles to handle the leading edge.  The area between the body and the control surfaces being exceptionally tricky.

Can they make a special tile type that totally covers the leading edge on one side and keys in with the hexagon tiles on the other? If the curvature of the wings was constant it would only require one such tile... that said I'm thinking the curvature won't be the same will it? in which case it means unique tiles all along leading edge or at least on the variable curvature bits...

[Eylrid]

Having a large number of intricate tile shapes might not be a problem by itself as long as they don't require a manual repairs. Unlike on the Shuttle there is no chance of getting hit by debris during launch.

For tiling a rounded cone they could do it using the same slightly warped hexagon at each height point.

But they might eventually go for unique shapes of varying thickness anyway in order to optimize mass.

It's actually mathematically impossible to tile a rounded cone with any sort of uniform warped hexagon.  If you try it, you'll find that it works alright at the beginning, but as the surface continues curving, your hexagons will become more and more distorted until you're forced to add pentagons to the mix.  This is a direct consequence of the Euler Identity.

There's just no way to avoid requiring a large number of different tile shapes when trying to tile a surface with non-zero curvature.  (A cylinder has, mathematically speaking, zero curvature, just to be clear on this.  Not so for a rounded cone)

Mathematically you're correct.

But engineers like to cheat

1 type of tile in 2 forms, whole and cut in half.

PS: I'd like to emphasize that only 200deg of SS needs to be tiled.

Long straight lines like that are problematic.

In the words of Elon when asked why hexagons:

No straight path for hot gas to accelerate through the gaps

https://twitter.com/elonmusk/status/1107379727302451200

All of the verticals are the same size as the standard hex tile and the other lines are shorter.

I wasn't refuting your idea, but DusanC's, which has long horizontal cuts: https://forum.nasaspaceflight.com/index.php?topic=50620.msg2073315#msg2073315. I cross quoted from the prototype thread to this one where the discussion is on topic.

As for your idea, I think it's one of the better ones. It seems like a good compromise between covering the curved surface and keeping the number of unique shapes down.

[livingjw]

Then the tiles were glued to thin nylon fabric layers (AIU like the stuff womens tights are made out of) to handle the contraction and expansion.

In more detail:

- Shuttle Tiles are attached using high temperature silicone RTV adhesive. Similar to high temperature silicone adhesive from auto store.

- Tiles are bonded to a ~4 mm thick Strain Isolation Pad, SIP, made from Nomex felt. Nomex is related to Nylon.

- The SIP is, in turn, bonded to the aluminum Shuttle Skin with another layer of Silicone RTV adhesive.

John

[LF]

BTW

A few years ago DLR ran a sounding rocket test with the Brazilian space agency.

They built a nose cone out of standard sized RCC tiles, mostly triangles, pentagons etc.

The key development was the software to confirm despite the segments being flat that the aerodynamics were OK and the flow was smooth enough to avoid turbulent flow (which multiplies heat transfer between 4x and about 6x).

Oh, this seems super super cool -- I was googling to try to learn more, and to clarify, are you talking about these tests or something else? https://en.wikipedia.org/wiki/Sharp_Edge_Flight_Experiment

From looking at it, it makes me think that the most obvious applications (and maybe obvious reasons it was funded) are hypersonic missiles, and that one of the hard jobs a hypersonic missile would be expected to do (zipping through atmosphere at high speed without burning up) can be aerodynamically similar to lobbing something nearly into orbit and having it re-enter.

It also makes me wonder how much interesting stuff is out there that we can read and draw conclusions about. Some cursory googling brings up stuff not much more interesting than the wikipedia entry.