I need to develop a model to predict the orbital decay time for my graduation project but I searched for data that can help in that project and I failed to obtain such data.
Can any one suggest a useful data set?
Asked
Active
Viewed 97 times
2
astrosnapper
- 2,506
- 1
- 13
- 24
amr ehab abdelhamed
- 21
- 1
-
These are different but related - there might be something with which to start a search How might one approach using AI (convolutional neural network) to predict collisions in orbit? and Does Privateer's tracking software Crow's Nest actually offer users something better than the previous state of the art? – uhoh Oct 27 '23 at 19:26
-
and also What does NASA plan to do with the world's largest quantum computer? and recent follow-up What DID NASA do with "the world's largest quantum computer"? (re question from 2013) – uhoh Oct 27 '23 at 19:34
-
2@uhoh - Although it's four years old now there was an answer to one of your questions by a data specialist who said, "The sort of problems that neural networks are amenable to solve tend to be rather soft and fuzzy (both in the mathematical sense and the is-this-picture-a-cat? sense). Physics simulation problems have well-defined rules, and the best way of solving them is usually through good old fashioned brute-force simulation." Have things changed since then? Maybe one of the links to your other questions discusses this, I haven't had a chance to read through all of them yet. – Steve Pemberton Oct 28 '23 at 13:22
-
@StevePemberton I suppose you could call breakthroughs like AI winning at Chess and Go solving "fuzzy" problems, more recent solutions like molecular structures demonstrate that we may need to forever get rid of the preconceived notion that brute force numerical solutions are the ONLY way to solve physical problems. cf. ‘New era in digital biology’: AI reveals structures of nearly all known proteins and... – uhoh Oct 28 '23 at 23:37
-
@StevePemberton ‘The entire protein universe’: AI predicts shape of nearly every known protein and ‘The game has changed.’ AI triumphs at protein folding and NEWS 30 November 2020 ‘It will change everything’: DeepMind’s AI makes gigantic leap in solving protein structures – uhoh Oct 28 '23 at 23:40
-
@StevePemberton even more quantitatively see all the questions tagged machine-learning in Matter Modeling SE and in Chemistry SE and questions including Machine Learning in Physics SE – uhoh Oct 28 '23 at 23:48
-
@StevePemberton Numerical initial value problem solvers are really, really dumb; you have to go all the way to the bitter end. AI recognizes patterns. Currently it may not tell you the closest approach to the nearest meter, but it may end up being much better at flagging which objects to follow closely, which ones need better optical or radar observations, or which ones are likely to require that you move your spacecraft for next month. We're going to have to just wait and see how this unfolds ("folding" pun intended) – uhoh Oct 28 '23 at 23:48
-
@uhoh - I suppose one test would be to take position data for some asteroids and space vehicles from say twenty years ago, then run it through both AI and regular prediction calculations and see which one's results are closer to the actual current positions of those objects – Steve Pemberton Oct 29 '23 at 10:43
-
@StevePemberton if you cast the problem as one of numerical accuracy for a limited number of objects, numerical integration is still the tool of choice. At the moment, where AI shows potential is in situation where raw numerical evaluations are suboptimal (like Chess, Go, protein folding, many-body problems) AI doesn't replace numerical integration, it complements it. – uhoh Oct 29 '23 at 23:23
-
The physics calculation is straightforward given the orbital elements, the ballistic coefficient, and the atmospheric density profile. The orbital elements and ballistic coefficient are stable properties of the individual object, except right at the end of its life. The thing that's difficult to predict is the atmospheric density profile for low Earth orbits: it depends on the "space weather". There's plenty of public data on space weather. That's where AI might have an impact, but note that there is much scientific competition in this area: making an original contribution will be hard work. – John Doty Oct 30 '23 at 15:41
-
@JohnDoty "I need to develop a model to predict the orbital decay time for my graduation project..." it sounds like a learning experience, rather than "making an original contribution" may be the primary goal in this case. – uhoh Oct 30 '23 at 21:46