When Does OpenMC Stop / Adjusting Accuracy of the Result?

User Support
1

Hi Everyone,

I’m working in one of my classes using OpenMC as a benchmark. My professor mentioned that since the program uses a Monte Carlo method, the program should stop once it has reached a solution that is close to the actual solution.

Monte Carlo methods are new to me, so I’ve been reading about how Monte Carlo methods work, looking at the simple case of calculating the probability of a coin toss landing heads ten times in a row and how if you ran the experiment a million times you would have a certain percentage of heads in a row. I’m trying to understand how many trials OpenMC runs. I’ve read through the documentation but I don’t understand when the program terminates and if there’s a way to limit the number of runs to increase speed but maintain an accurate result.

Could someone please help me understand how OpenMC determines when to stop?

Thank You

2

Hi Blake,

OpenMC normally requires that the user specify how many batches of particles to simulate and how many particles are in each batch. Thus, the product of the two will determine how many particles in total are simulated and hence how long the simulation will take (directly proportional to the number of particles). Note that this doesn’t tell you anything about the statistical uncertainty on any tallied quantities, so the most naive thing to do is to perform a short simulation, see where the uncertainty is at the end, and then project how many batches would have been required to reach a desired level of uncertainty. OpenMC also has a feature called “tally triggers” where you can indicate that the simulation should stop once a specified level of uncertainty is reached.

Hope this helps and let us know if you have further questions!

Best,
Paul

3

Oh, I see. I also had another question about the output. I read that k-effective value is the average number of neutrons from one fission to cause another fission. But I couldn’t find how this is related to the output OpenMC gives. OpenMC provides: k-effective[Collision, Track-length, and Absorption]. Is there a relation to the k-effective value I mentioned or is it entirely a different concept?

Thanks!

4

Yes, it is the same concept. OpenMC gives you several different estimates for k-effective as well as a “combined” estimate that usually has the lowest variance.

Best,
Paul