Re: [fluka-discuss]: correctness of simulation

From: <me_at_marychin.org>
Date: Sat, 11 Feb 2017 14:14:50 +0800

Dear Samsul,

Correctness -- are you concerned about correctness on the user's part or
correctness on the code's part?

Indeed I have seen 'anything under the sun' emerging from Monte Carlo
simulation of radiation interactions. Not from FLUKA though. FLUKA is
relatively conservative and is as safe as can be.

If we look at the vast range of projectile, energy, target and ejectile
that FLUKA simulates, we immediately recognise some domains which are
better understood / characterised / measured than others. Some domains
have been benchmarked -- please refer to FLUKA's biography. Benchmarked
against experimental data.

There are many other domains which failed to get to the funding bodies',
and therefore the experimenters', wish list. There are also domains
where experimental facilities are not yet available.

If we look at EXFOR, many data points scatter wildly everywhere without
convergence. If we look at TENDL, data points can be nice and clean and
shiny, but generated purely from models.

Inherent limitations aside, basics such as correctness in terms of
conservation of energy, momentum, A and Z -- we can put our fingers to
the keyboard, check and see with our own eyes and convince ourselves.

Simple irradiation conditions like an isotropic point source started
from the centre of a sphere. Set it to vacuum, we know the fluence a
priori. Set it to a homogeneous material, we know the energy deposition
and fluence a priori. This I find is the best way to gain faith and
confidence whether it is on the user's part or the code's part.
Sometimes it's just a matter of successful/unsuccessful relay of the
user's intention and the parser's (mis-)understanding.

Inter-code comparison is not benchmarking. Very few is credible. Most
tells us next to nothing about correctness.

:) mary

On 2017-02-10 22:06, Mattias Lantz wrote:
> Dear Samsul,
>
> As Viktor writes there is no standard check that you have done things
> right. Read carefully the beginner's manual, look at the relevant
> parts in the main manual (and also the course material that can be
> found on the Fluka web site, there are many good tips there).
>
> Then, instead of starting to build a complex geometry, it may be good
> to first try with simple examples where the output is not so complex.
> If the results from simple tests make sense with respect to what you
> can expect from known physics, then you have probably done well so far
> and can add more complexity to the simulation. This may of course not
> be so easy, depending on what kind of simulation you are running, but
> I find it to be a very useful approach for some of my simulations.
>
> Best wishes,
> Mattias Lantz
>
> On 2017-02-10 13:05, Віктор Родін wrote:
>
>> Dear Samsul,
>> Philosophical question. FLUKA gives us things which we created
>> ourselves .You can try to compare your results in other simulation
>> programs.
>> Best regards
>> Viktor
>>
>> 2017-02-10 10:15 GMT+02:00 <samsul_at_barc.gov.in>:
>>
>>> dear fluka experts,
>>>
>>> Is there any way or method in fluka by which one can test his
>>> simulation for correctness ?
>>>
>>> Regards,
>>>
>>> Samsul Arefin
>>>
>>>
>>
> __________________________________________________________________________
>>> You can manage unsubscription from this mailing list at
>>> https://www.fluka.org/fluka.php?id=acc_info [1]
>
>
>
> Links:
> ------
> [1] https://www.fluka.org/fluka.php?id=acc_info

__________________________________________________________________________
You can manage unsubscription from this mailing list at https://www.fluka.org/fluka.php?id=acc_info
Received on Sat Feb 11 2017 - 08:22:42 CET

This archive was generated by hypermail 2.3.0 : Sat Feb 11 2017 - 08:22:46 CET