Re: [Orekit Developers] issue 324 is probably a serious ugly bug

[Luc Maisonobe <Luc.Maisonobe@c-s.fr>]

Hi all,

Le 24/05/2017 à 23:34, paulcefo a écrit :

> I note that Rick Lyon's MIT Masters Thesis Report "Geosynchronous Orbit
> Determination using Space Surveillance Network Observations and Improved
> Radiative Force Modeling" (2004) has in Appendix E-One Panel Aerodynamic
> Force Model a consideration that includes atmosphere lift.


Le 25/05/2017 à 15:43, Ward, Evan a écrit :

> In Section 8.2 in [1] Hughes examines the different interactions the
> rarefied atmosphere can have with a spacecraft panel. He agrees with
> you on the cosine squared term for specular reflections. He also states
> (without providing a reference or much explanation) that in most
> interactions the gas particle is absorbed by the surface and then
> diffusely emitted latter at a much lower velocity. Ignoring the
> particles leaving the surface in diffuse reflection there is only one
> cosine term (due to the shrinking cross sectional area) and the
> resultant force is always along the relative velocity vector. So I think
> class is correctly implementing a multi-plate model assuming that the
> surface absorbs all of the incoming momentum from an air particle. 
> 
> In practice detailed drag modeling is hampered by the fact that we're
> really bad at predicting the atmospheric density. Perhaps a more
> detailed model would provide a coefficient for absorption vs. reflection
> incoming particle, but how much of a difference it makes is going to be
> situation dependent given all of the other uncertainties.
> 
> [1] Hughes, Peter C. Spacecraft Attitude Dynamics. Dover Publications.
> 2004.


OK. Following these two references, I will therefore implement a
combined model where the user will be able to set up coefficients
as desired. For most people, it will be consistent with previous
model, i.e. they will input CD for drag and CL for lift will be
defaulted to 0. I will use coefficients as close to Lyon's thesis as
possible, still ensuring the CD is consistent with what we used up
to now, so exixting users do nat experience any regression. If users
explicitly set up a non-zero CL, it will be used. Esimating either or
both parameters will be possible in orbit determination.

best regards,
Luc