Dear Masoomeh,
before implementing bias, I would suggest you some improvement in your input file:
- photons and electrons produced by 10 MeV electrons interaction on copper will be easily stopped by the concrete so you need to increase
significantly the the number of primaries to several million. You simulation run through in a 40-50 seconds (0.5 ms/primaries) on a single core and I would suggest you to run for a few hours just to get an idea;
- limit a bit the extension of your simulation. The external VOID, to which you assigned AIR as material, has an AIR-AIR interface along the Y-axis where particles will be tracked (this is also CPU time). This will save you about 10% CPU time (I tested your case);
- the binning you are scoring with is very fine. When computing the equivalent dose after so much shielding you can get a coarser binning (both by live-rebinding with flair of just use a second coarser USRBIN.
Coming to the biasing:
- Biasing need an understanding of the evolution of the particle spectra thought the material you want to bias;
- The importance biasing you are using is a big aggressive.4->32->128->512->1024;
- You are biasing over concrete, the distances between the different importance regions should compensate the absorption of the particle thought it (which are also getting less and less energetic);
- You can reduce the importance of regions where you are interested to save CPU time. Also this can make you save some CPU time.
In conclusion:
- Observe the dose on a larger BINs, for example by rebinning the USRBIN you already have;
- Improve the segmentation of the biasing;
- Do a few runs., for example 10e5, 10e6 (40m on 1 core), 10e7(6 hrs on 1 core), and check the evolution of the error in the region you are interested in to evaluate which is the number of primaries you need to reach your goal;
- Spawn the job to profit from multicore CPUs and adjust the number of primaries accordingly.
Best,
Vittorio
> On 06 Mar 2016, at 14:15, Masoomeh Yarmohammadi <myarmohammadi_at_ipm.ir> wrote:
>
> Dear All,
>
> I have a question about reducing the statistical error. For reaching to this goal I used the Biasing card between regions. But the error is still more than 10% for the last layer (region) in the forward shield wall. It puts me on doubt about the value of does for that region (with more that 10% error). The biasing that I used is Correct? My doubt about does EQ rate value of this high percent error regions (specially, after the wall) is correct? How can I reduce this error for that region?
>
> The files are attached. This is for a shield of a continuous 12uA, 10MeV Electron linac beam which hit a 3mm copper target. I normalized the dose output data with 27e10 in plotting with flair to convert it from pSv/primary to uSv/h. This make me able to compare the does rate after the shield with the safe value of 10uSv/h.
>
> The last point, I implemented that Biasing for the Forward wall (WF group) as the first step and my question is around that part and after that.
>
> The comments and suggestions are absolutely welcome.
>
> Many thanks
> Masoomeh Yarmohammadi Satri
> Post-Doctoral Research Fellow, School of Particles and Accelerator, Institute for Research in Fundamental Sciences (IPM)
> E-mail : myarmohammadi_at_ipm.ir, myarmoha_at_cern.ch
>
> <Does-Eq-with Biasing-RR-Rigion 1-2-Norm to 27e10-uSv per h-End of WF5.png><Error-with Biasing-RR-Region1-2-WF5.png><Bias-Main-v1-3mm-target-gamma-elec-energy-spec.flair><Bias-Main-v1-3mm-target-gamma-elec-energy-spec.inp>
__________________________________________________________________________
You can manage unsubscription from this mailing list at
https://www.fluka.org/fluka.php?id¬c_info
Received on Sun Mar 06 2016 - 19:16:08 CET