Modeling Atmospheric Re-entry & Determing Optimal Re-entry Parameters

[roldy]

I'm taking a course in adaptive neural networks and became interested in trying to see if I can use neural networks to find the optimum values for the control parameters in atmospheric re-entry in order to reach a specified target on Earth. I've come across a paper that discusses this and goes into some detail. First and foremost, I need to determine the bounds on the re-entry flight, i.e max and min entry angle to avoid skip-out and excessive g-forces.

I have the equations of motion setup as well as penalty functions for altitude and for pilot g-force loading.

The steepness of the entry angle is determined by increasing the entry angle from 90 to 180 in small steps and solving the pilot penalty minimization problem. The upper bound is the largest entry angle that yields a pilot penalty of no more than 1. I am getting different results from what is shown in the paper.

Procedure:

The program takes initial values for the state variables and uses them in the numerical integration. Numerical integration is stopped when the difference (1 - pilot penalty) <= 0.01 and when the velocity <= Mach 2. This is repeated for the next theta value. I am increasing theta by 1 degree and the time steps by 0.01.

Once I have all the entry angles for the pilot penalty minimization, I then search for the largest of these. This will then be the upper bound.

The files for this project as well as the paper if anyone is interested are here http://www39.zippyshare.com/v/22155505/file.html. The section in the paper that I am working on right now are pages 15-29. If someone could take a look at my program and see if I messed up anywhere that would be awesome.

 

[eljose79]

Hello,

It is great to see your interest in using neural networks for optimizing control parameters in atmospheric re-entry. I have taken a look at your program and the paper you have shared. Firstly, let me say that your approach seems sound and well thought out. However, I do have a few suggestions that could potentially improve your results.

1. Consider using a more sophisticated numerical integration method such as Runge-Kutta instead of the simple Euler method. This can help improve the accuracy of your results.

2. Instead of increasing the entry angle by just 1 degree, you can try using a smaller step size (e.g. 0.1 degrees) to get more data points and a smoother curve. This can also help in finding the optimal upper bound more accurately.

3. It might be helpful to also consider the effect of other control parameters, such as the angle of attack and lift coefficient, on the pilot penalty. This can give you a more comprehensive understanding of the system and potentially lead to better optimization results.

4. Have you considered using a neural network to directly optimize the control parameters instead of using a numerical integration method? This could potentially lead to faster and more accurate results.

I hope these suggestions are helpful to you. Keep up the good work and I wish you success in your project!