Author Topic: SpaceX Falcon Mission Simulations  (Read 25152 times)

Online Flying Beaver

Re: SpaceX Falcon Mission Simulations
« Reply #40 on: 04/12/2017 09:43 PM »
Been playing around in Flight Club recently (Getting to orbit with a F9 from scratch on the first attempt was pretty surreal, thanks KSP!) and I've come up with this rather interesting profile.

10,000kg payload to 32.5deg 250km by 220km orbit with second stage burn to depletion.

MECO at t+145s, 1.5km/s.

First stage boostback to LZ-1, with only 99km apogee.

Extremely low entry force on S1 (only slightly higher than MaxQ on ascent), with extended (20s) entry burn.

Seems like a pretty nice time for the first stage. Are there any obvious issues with it though?

https://www.flightclub.io/results/?id=0d9a43bc-328a-481d-9cbc-22321c3f95d3&code=NONE
« Last Edit: 04/12/2017 09:44 PM by Flying Beaver »
Saw OG-2 Booster Land in person 21/12/2015.

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #41 on: 04/13/2017 12:35 PM »
...10,000kg payload to 32.5deg 250km by 220km orbit with second stage burn to depletion.
MECO at t+145s, 1.5km/s.
First stage boostback to LZ-1, with only 99km apogee...

Seems like a pretty nice time for the first stage. Are there any obvious issues with it though?

CRS-10 was the most recent RTLS profile: MECO at T+143s, 1.675km/s. The payload was about 7,000kg, but if I re-run my CRS-10 sim with your payload, I only see a reduction of about 25m/s at MECO. Are you sure about your 1.5km/s value, two seconds later? What throttle settings are you using?

Re: SpaceX Falcon Mission Simulations
« Reply #42 on: 04/13/2017 01:35 PM »
https://www.flightclub.io/results/?id=0d9a43bc-328a-481d-9cbc-22321c3f95d3&code=NONE

If you visit the link posted, you'll see there's a graph of throttle settings over time. The MaxQ throttle looks like 70% from T+50 to T+75. Bit extreme, I reckon, but probably explains the dV deficiency.

Seems like a pretty nice time for the first stage. Are there any obvious issues with it though?

The main problem I have is that your landing burn lasts for almost 90s. The retropropulsion through the entire lower atmosphere is the reason why the aerodynamic pressure on descent is so low, but this is extremely inefficient! Also the landing burn is at about 50% throttle the entire time which is also very inefficient. Best to do as short a landing burn as possible to minimize gravity losses - so you wanna maximise aerobraking and burn at a high throttle setting when you do start the burn.

Kudos getting to orbit on the first attempt though, that's pretty impressive :)

Online Flying Beaver

Re: SpaceX Falcon Mission Simulations
« Reply #43 on: 04/13/2017 09:48 PM »
https://www.flightclub.io/results/?id=0d9a43bc-328a-481d-9cbc-22321c3f95d3&code=NONE

If you visit the link posted, you'll see there's a graph of throttle settings over time. The MaxQ throttle looks like 70% from T+50 to T+75. Bit extreme, I reckon, but probably explains the dV deficiency.

Seems like a pretty nice time for the first stage. Are there any obvious issues with it though?

The main problem I have is that your landing burn lasts for almost 90s. The retropropulsion through the entire lower atmosphere is the reason why the aerodynamic pressure on descent is so low, but this is extremely inefficient! Also the landing burn is at about 50% throttle the entire time which is also very inefficient. Best to do as short a landing burn as possible to minimize gravity losses - so you wanna maximise aerobraking and burn at a high throttle setting when you do start the burn.

Kudos getting to orbit on the first attempt though, that's pretty impressive :)

But it is still 10tons to LEO, with RTLS, no matter how inefficient the landing burn.

Would the reason for SpaceX not using a profile like this be about having a buffer for second stage performance?
Saw OG-2 Booster Land in person 21/12/2015.

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #44 on: 04/13/2017 11:34 PM »
But it is still 10tons to LEO, with RTLS, no matter how inefficient the landing burn.

Would the reason for SpaceX not using a profile like this be about having a buffer for second stage performance?

That, and if you are launching Dragon to the ISS (which is probably why you can afford RTLS), your apogee should be at least 400kms, and your orbital inclination around 51.64.

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #45 on: 05/02/2017 12:15 PM »
A preliminary S1 profile for NROL-76. Thrust still appears to be 90% of rated, so perhaps not Block 4.

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #46 on: 05/02/2017 12:19 PM »
I already did analysis here and here, if you want to compare and contrast.

Offline Semmel

  • Full Member
  • ****
  • Posts: 1112
  • Germany
  • Liked: 793
  • Likes Given: 2263
Re: SpaceX Falcon Mission Simulations
« Reply #47 on: 05/02/2017 12:25 PM »
A preliminary S1 profile for NROL-76. Thrust still appears to be 90% of rated, so perhaps not Block 4.

Interesting. Did you subtract the acceleration caused by gravity? Also, what is the jump in velocity near 245s and 337s?

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #48 on: 05/02/2017 12:35 PM »
I already did analysis here and here, if you want to compare and contrast.

Thanks, I should explain what I've done, perhaps it's slightly different. Using the telemetry as a guide, I've created a simulation of the launch profile. Then I've output the velocity, altitude and acceleration from the sim to a spreadsheet. So, the acceleration is new information. Other parameters like thrust, orientation, dynamic pressure and heating are also available, but I'm still refining this particular sim, there was so much new information in the broadcast!

Interesting. Did you subtract the acceleration caused by gravity? Also, what is the jump in velocity near 245s and 337s?

Yes, relative acceleration. The two jumps are an artefact of the sim. Above 150kms in altitude it considers the velocity to be no longer including the earths rotation. Well spotted!

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #49 on: 05/02/2017 12:57 PM »
I already did analysis here and here, if you want to compare and contrast.

Thanks, I should explain what I've done, perhaps it's slightly different. <snip>

Ah yes, didn't realise that, good point. Mine are just straight plots from the telemetry.

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #50 on: 05/02/2017 01:05 PM »
Mine are just straight plots from the telemetry.

You've also derived acceleration though, and very interesting to see the noise in that data. It is quite a challenge working out what information to discard.

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #51 on: 05/02/2017 01:33 PM »
Yes, the acceleration data is quite noisy, when going down to single second resolution. It might be possible to get better data out of the stream, as the velocity data updates multiple times a second (altitude data is much coarser). Conceivably, you could take multiple measurements per second and average them out, to smooth out the noise a bit. That's a lot of work though, and I don't know what it'll add. We're already working off figures that have gone through a lot of filtering and lag etc before they hit the stream, so the accuracy is going to be limited regardless.

Offline manoweb

  • Full Member
  • **
  • Posts: 204
  • Tracer of rays
  • Hayward CA
  • Liked: 85
  • Likes Given: 83
Re: SpaceX Falcon Mission Simulations
« Reply #52 on: 05/02/2017 06:04 PM »
How do you guys extract data from the webcast? Perform an OCR frame by frame to extract the speed and altitude?

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #53 on: 05/02/2017 06:12 PM »
I literally sit there and copy the numbers off manually. Every 5 seconds for the whole flight, every second for the burns.

Online OneSpeed

  • Full Member
  • ****
  • Posts: 536
  • Liked: 557
  • Likes Given: 451
Re: SpaceX Falcon Mission Simulations
« Reply #54 on: 05/02/2017 09:45 PM »
I literally sit there and copy the numbers off manually. Every 5 seconds for the whole flight, every second for the burns.

Similar. I then re-run the sim many times tweaking throttle, orientation and payload until the data points line up as close as I can get. It's interesting how often this reveals constant throttle settings for extended periods. The simulator is then effectively applying a very strong filter to the noise in the source data.

Offline manoweb

  • Full Member
  • **
  • Posts: 204
  • Tracer of rays
  • Hayward CA
  • Liked: 85
  • Likes Given: 83
Re: SpaceX Falcon Mission Simulations
« Reply #55 on: 05/02/2017 10:33 PM »
I did the automated OCR in the past for a whole different application. I will try this thing...

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #56 on: 05/03/2017 06:55 AM »
Sounds like it could be very useful.  I'd love to hear how you got on.

Offline manoweb

  • Full Member
  • **
  • Posts: 204
  • Tracer of rays
  • Hayward CA
  • Liked: 85
  • Likes Given: 83
Re: SpaceX Falcon Mission Simulations
« Reply #57 on: 05/03/2017 08:17 AM »
If the system allows the attachment, I have created a CSV file with the data. It's completely raw and it does NOT have a timestamp, just velocity and altitude data, for each line. Timestamp can be inferred as each frame is 1/29.970 of a second (don't get me started). Please suggest a format for the timestamp I can add it.


Offline manoweb

  • Full Member
  • **
  • Posts: 204
  • Tracer of rays
  • Hayward CA
  • Liked: 85
  • Likes Given: 83
Re: SpaceX Falcon Mission Simulations
« Reply #58 on: 05/03/2017 08:31 AM »
How I did it:
- download the Youtube video of the launch. It's probably illegal/against the terms of use of Youtube in some jurisdictions.
- inspect the file to find T start and T end of the flight
- cut it out with ffmpeg:
ffmpeg -ss 11:59 -i nrol.mp4 -t 9:04 -vcodec copy -acodec copy cut.mp4
- crop out the relevant data on screen and save individual images of each frame:
ffmpeg -i cut.mp4 -filter:v "crop=192:24:1040:140" img/frame%05d.png
- at this point we need to enhance the images so that the OCR program has an easier time decodign them: I figured that pushing the levels and inverting the image would help:
for i in *.png; do convert $i -threshold 40% -negate proc/$i; done
- now, gocr works better if instead of giving individual images, we group several, so it can create some sort of internal database and other second order effects; however, we have more than 16k files and if we really put them together we get a huge image and gocr will runaway. So I decided to analyze 1000 images at a time. Also, by specifying we want only digits and the decimal point, and lowering the confidence threshold, we get an acceptable result:
for i in `seq -w 0 16`; do echo $i; convert -append frame${i}*.png table.png && gocr -a 80 -C "0123456789." table.png >> list.txt; done
- at this point I had to manually correct a bunch of non-recognized characters "_" but it just took few minutes
- create the final CSV:
while read -r a b; do echo "$a,$b"; done <list.txt >NROL76.txt


All of the above done in pretty much any classic Linux environment with bash gocr ffmpeg and imagemagick
« Last Edit: 05/03/2017 08:37 AM by manoweb »

Offline Welsh Dragon

  • Full Member
  • **
  • Posts: 210
  • Liked: 206
  • Likes Given: 26
Re: SpaceX Falcon Mission Simulations
« Reply #59 on: 05/03/2017 10:48 AM »
Cool, seems to match well with the data I got. Be interesting to look at the burns in more detail, to see whether averaging frames can help bring the noise down a bit.

Tags: