sorry I forgot:
the random seed does not influence statistics, HOWEVER one must ensure
that "parallel" runs on different CPUs have different random seeds.
Paola
On 6/16/21 9:09 PM, Paola Sala wrote:
>
> Dear Zhiyuan,
>
> There is no general number of particles, it depends on the problem
> under consideration.
>
> The number of cycles should be at least 5 because the statistical
> errors are evaluated by Fluka when the results from different cycles
> are combined, at the end of the runs: the variance is calculated from
> the weighted variance among all different cycles from all runs of the
> same input . This procedure has no meaning if the number of cycles is
> too small.
>
> The driving quantity for statistics is the total number of events, the
> product Nt = N * Nr *Ns
>
> The random seed does not influence at all the statistical error.
>
> This said, the actual value of the statistical error depends on how
> many times a detector receives a signal.
>
> No signal means no statistics even after an infinite number of primaries.
>
> In your case, the 250 MeV proton beam stops a few centimeters before
> the end of the target.
>
> Products of proton nuclear interactions will stop before or slighly
> after the proton range, within the target, except for neutrons.
>
> Detectors placed downstream the target, will "see" only the few
> neutrons and the even fewer charged hadrons produced by neutron
> reinteractions. I doubt that you will ever be able to achieve some
> statistical significance on charged hadrons, even using biasing.
>
> Regards
>
> Paola
>
>
>
>
>
> On 6/9/21 5:33 AM, 梅志远 wrote:
>>
>> Dear experts,
>>
>> I recently ran a simulation by Fluka. Unfortunately, I found the
>> statistical errors of the results (the concerned particles' flux
>> scoring by the USRTRACK cards)were very big!
>>
>> I have tried many ways to reduce the errors. However, the drop of the
>> errors are unapparent. I have some questions about the statistical
>> errors of calculating results:
>>
>>
>> 1) How many primary particles should I set in the START card to
>> obtain a reasonable result? I have tried to set the primary particles
>> equal to 1E8. However, the errors are still unacceptable.
>>
>>
>> 2) If the number of primary particles in the START card is N, the
>> number of cycles to run is Nr, and the number of spawned jobs is Ns,
>> I found the number of 'total' primary particles in the result filesNt
>> = N * Nr *Ns. So, what is the most crucial numbers influencing the
>> statistical errors, N or Nt?
>>
>>
>> 3) The Fluka manual says that the cycles of run should be 4 or 5 to
>> obtain the statistical errors. If I run the FLUKA project by a
>> spawning ways, the 'total' cycles of run should equal Nr * Ns. I
>> really want to know how should I set the cycles of run to be 4 or 5.
>> Should I directly set Nr = 4, or should I set Nr * Ns = 4? How dose
>> the number of spawned jobs Nsaffect the statistical errors?
>>
>>
>> 4)Do the random number seeds in RANDOMIZ cards affect the statistical
>> errors of the results? Should I adopt different seeds in different
>> cycles of run to obtain a more reasonable results? If I should, how
>> can I implement it.
>>
>>
>> I have attached my inp files here.
>>
>>
>> Any help is appreciated!
>>
>>
>> Best regards
>>
>> Zhiyuan
>>
>>
>> --
>>
>> Zhiyuan Mei
>> meizhiyuan_at_hust.edu.cn
>>
>> /---------------------------------------------------
>> Institute of Applied Electromagnetic Engineering(IAEE)
>> Huazhong University of Science and Technology(HUST)
>> Address: Luoyu Road 1037, Wuhan, China
>> Tel: +86-155-2701-7168
>> /---------------------------------------------------
>>
>>
>>
__________________________________________________________________________
You can manage unsubscription from this mailing list at
https://www.fluka.org/fluka.php?id=acc_info
Received on Wed Jun 16 2021 - 22:48:35 CEST