Hey,
Is anyone else having a problem where transient simulation systematically estimate a keff lower than that estimated by a static simulation?
two things I've noticed:
1) The degree of underestimation is inversely proportional to the length of the time step. A very short time step (the tmax [last] entry on the tme card) creates a very large underestimation whereas a large time step creates a small under estimation
2) The degree of underestimation is inversely proportional to the number of particles. A situation with many particles has a keff closer to the SS value.
Any clues as to what's going on?
Transient Simulation Systematically Underestimates Keff

 Posts: 455
 Joined: Fri Sep 07, 2012 1:43 pm
 Security question 1: No
 Security question 2: 92
Re: Transient Simulation Systematically Underestimates Keff
Hi,
the main problem with transient keff is that there actually is no such thing.
By definition, the keff is the eigenvalue of the keigenvalue formulation of the transport equation, which is not what we solve in transient simulations.
That said, Serpent does print out all manner of things that are named "KEFF" something in the _res.m file for transient simulations. The explanations for these things can be found in this post: http://ttuki.vtt.fi/serpent/viewtopic.p ... 2542#p7184
I think the best way to approach transient simulations would be to forget about keff completely. If you want to verify that the reactivity insertion in your transient model corresponds to the experimental value it is probably better to calculate and compare the data that was used to estimate the experimental reactivity insertion in the first place (i.e. timedependent neutron flux signal or reactor period or something similar).
However, you can calculate something that resembles a time dependent keff by using detectors with time binning:
The tallied variables are
nsf = fission neutron production (prompt + delayed)
capt = capture reactions
fiss = fission reactions
nuxn = neutron production in inelastic reactions (n, xn), (n, alpha + n) etc.
leak = neutron leakage
Then, using these tallied variables the "multiplication factor" is calculated as
k = production/loss
or
k = nsf/(capt + fiss  nuxn + leak)
The inelastic production is subtracted from the loss terms.
The different variables can be tallied with different response functions for detectors:
nsf = dr 7
capt = dr 2
fiss = dr 6
nuxn = dr 16
leak = dr 17
You should note that this "multiplication" factor might not equal the timeindependent eigenvalue mode multiplication factor at all points.
Ville
the main problem with transient keff is that there actually is no such thing.
By definition, the keff is the eigenvalue of the keigenvalue formulation of the transport equation, which is not what we solve in transient simulations.
That said, Serpent does print out all manner of things that are named "KEFF" something in the _res.m file for transient simulations. The explanations for these things can be found in this post: http://ttuki.vtt.fi/serpent/viewtopic.p ... 2542#p7184
I think the best way to approach transient simulations would be to forget about keff completely. If you want to verify that the reactivity insertion in your transient model corresponds to the experimental value it is probably better to calculate and compare the data that was used to estimate the experimental reactivity insertion in the first place (i.e. timedependent neutron flux signal or reactor period or something similar).
However, you can calculate something that resembles a time dependent keff by using detectors with time binning:
The tallied variables are
nsf = fission neutron production (prompt + delayed)
capt = capture reactions
fiss = fission reactions
nuxn = neutron production in inelastic reactions (n, xn), (n, alpha + n) etc.
leak = neutron leakage
Then, using these tallied variables the "multiplication factor" is calculated as
k = production/loss
or
k = nsf/(capt + fiss  nuxn + leak)
The inelastic production is subtracted from the loss terms.
The different variables can be tallied with different response functions for detectors:
nsf = dr 7
capt = dr 2
fiss = dr 6
nuxn = dr 16
leak = dr 17
You should note that this "multiplication" factor might not equal the timeindependent eigenvalue mode multiplication factor at all points.
Ville

 Posts: 1
 Joined: Mon Jun 18, 2018 10:10 pm
 Security question 1: No
 Security question 2: 93
Re: Transient Simulation Systematically Underestimates Keff
Sorry to resurrect an old thread, but I would really appreciate some clarification on the MT numbers here (for myself and anyone else who finds this thread later on).
2. Conversely, the wiki, the Serpent 1 manual and the ENDF6 document all say that total fission is MT=18. Using that in a detector however, returns the following error:
3. According to the wiki, dr 17 points to the (n,3n) reaction. To estimate the leakage I have been using a surface current detector (ds) to get the net current through the outermost surface. Does this make sense?
4. Should I also be getting the inelastic production (MT=4?) to subtract from the losses?
Thanks a lot for your help.
EDIT: 5. Kind of off topic, but I've been questioning whether dr 17 actually is (n,3n) because the numbers I've been getting back are two orders of magnitude larger than those from dr 16 (n,2n).
1. From what I can see in the ENDF MT page on the wiki, 6 and 7 aren't valid numbers. The ENDF6 manual referenced on the page says that MT=69 are not valid in ENDF6. Both of these actually work in Serpent 2.1.29 however.nsf = dr 7
capt = dr 2
fiss = dr 6
nuxn = dr 16
leak = dr 17
2. Conversely, the wiki, the Serpent 1 manual and the ENDF6 document all say that total fission is MT=18. Using that in a detector however, returns the following error:
Code: Select all
MT 18 not allowed in response function
4. Should I also be getting the inelastic production (MT=4?) to subtract from the losses?
Thanks a lot for your help.
EDIT: 5. Kind of off topic, but I've been questioning whether dr 17 actually is (n,3n) because the numbers I've been getting back are two orders of magnitude larger than those from dr 16 (n,2n).

 Posts: 455
 Joined: Fri Sep 07, 2012 1:43 pm
 Security question 1: No
 Security question 2: 92
Re: Transient Simulation Systematically Underestimates Keff
Hi,
a distinction should be made between positive and negative values for the detector response.
Positive values refer to ENDF reaction MT numbers (http://serpent.vtt.fi/mediawiki/index.p ... ion_MT.27s) and negative values refer to Serpent specific macroscopic reaction numbers (http://serpent.vtt.fi/mediawiki/index.p ... on_numbers).
Ville
a distinction should be made between positive and negative values for the detector response.
Positive values refer to ENDF reaction MT numbers (http://serpent.vtt.fi/mediawiki/index.p ... ion_MT.27s) and negative values refer to Serpent specific macroscopic reaction numbers (http://serpent.vtt.fi/mediawiki/index.p ... on_numbers).
Ville