Monte Carlo Simulation Error: NaN Time in System State Derivative

[luigigisolfi]

Question

User:
Hey everyone, I have a small question. I am currently running a Monte Carlo simulation and while running it I got the following error for one of the cases:
Error when computing system state derivative. Input time is NaN
I checked and the simulation starts fine and it runs for about 1.5 days before breaking down. I have not been able to figure out why it breaks down at some point. The MC is only about different initial translational states and all other cases are running fine. I have also checked the output and the propagation does not get stuck at a small time interval. Any ideas of what might be happening?
Thanks a lot!

[luigigisolfi]

Answer

Tudat-team:
Hi, without a look at the code, there's really not much we can say, except that the state derivative at some point is called with t=NaN. This is usually indicative of the step-size control algorithm going haywire. I would recommend:

• Have a look at the state/dependent variable history of the specific case that yields this issue. Use this to find out where the error comes from
• Omit the case(s) that yield this issue from the Monte Carlo analysis, but make sure to find out why it happens

[luigigisolfi]

User:

Interesting. I will ignore this result for now and take a look at the variables to see if I can find what is happening. Thanks a lot!

Quick question, Could it be possible that if I am using a strict termination condition
terminate_exactly_on_final_condition=True
and the termination condition is met and the simulation keep running for whatever reason, the t=NaN error could come up? I am asking this because when looking at the results from the MC that give me this error it seems that the termination condition has been violated.
I have a 120km altitude termination condition and when the error comes up I see that the termination altitude is a bit lower, around 119km or so.

[luigigisolfi]

Tudat-team:
That's definitely possible. Even if you set the strict termination condition, the propagation will first violate the termination condition, and once it's noticed that it's violated it, it will go into a root finding algorithm to converge to the exact termination condition

[luigigisolfi]

User:

Perfect, good to know!