DAKOTA

DAKOTA developed by Sandia National Laboratory represents a tool for optimization of designs, sensitivity and uncertainty analysis of virtually unlimited kinds of simulations. The user manual can be found at DAKOTA homepage.

Running DAKOTA

It is the best to illustrate DAKOTA function by an example. Running DAKOTA requires typically DAKOTA input file, template file to generate variants of input files for the actual calculation tool, and driver script that defines for DAKOTA how to read the results. This example is built for Serpent and analysis of multiplication factor uncertainty due to fuel salt density uncertainty.

DAKOTA input file defines calculation parameters. Especially the sampling method, default value of the analysed parameter and its standard deviation.

environment
 tabular_data
 output_precision 1e-16
 results_output

method
 sampling
   sample_type lhs
   seed = 52983
   samples = 100

model
 single
 
variables
 active all
 normal_uncertain = 1
   means           -2.06684099705179
   std_deviations  0.02
   descriptors     'saltDensity'

interface
 analysis_drivers = 'driver.sh'
   fork
     parameters_file = 'params.in'
     results_file    = 'results.out'
     file_save

responses
 response_functions = 1
 descriptors = 'kinf' 
 no_gradients no_hessians

Serpent template file with {saltDensity} keyword to be replaced by the variations.

set title "Graphite block 12 cm, salt channel 5 cm"

pin 1
salt     5
graphite 

pin e
graphite

lat lat.1  2  0.0 0.0 3 3 12
e e e
 e 1 e
  e e e

surf S.1 hexyc 0.0 0.0 6

cell C.1  0  fill lat.1 -S.1
cell C.O  0  outside     S.1

mat salt {saltDensity}
 3007.02c 3.60000E-01
 9019.02c 5.48000E-01
 4009.02c 9.00000E-02
92235.02c 1.01214E-04
92238.02c 1.89742E-03
92234.02c 9.04661E-07
92236.02c 4.63607E-07

mat graphite -1.8 moder grph12 6012
 6012.02c  0.9893
 6013.02c  0.0107
therm grph12 900 gr30.85t gr30.86t

set acelib "/opt/ENDFB-8.0/endfb80.xsdir"
set declib "/opt/ENDFB-8.0/sss_endfb80.dec"
set nfylib "/opt/ENDFB-8.0/sss_endfb80.nfy"

set bc 3

set pop 10000 500 20

The next file is driver.sh file that depends on the computational environment. The presented version works with the NEcluster. This driver.sh file defines preprocessing command for DAKOTA, check for existence of output files, submit command for the Serpent job, waiting for the results, and command to get results from Serpent output file.

#!/bin/bash

dprepro params.in msrG12S5_template msrG12S5_input

if [ -f msrG12S5_input.done ]; then
    rm msrG12S5_input.done
fi

qsub runScript.sh

while [ ! -f msrG12S5_input.done ]; do
  echo "#"
  sleep 30
done

awk '/IMP_KEFF/ {print $7,'kinf'}' msrG12S5_input_res.m > results.out

For completeness also the runScript.sh is given here. It produces at the end also the output file for which the driver.sh script searches.

#!/bin/bash
#PBS -V
#PBS -q gen5
#PBS -l nodes=4:ppn=4
#PBS -l mem=4000mb

module load mpi
module load serpent

#### cd working directory (where you submitted your job)
cd ${PBS_O_WORKDIR}

#### Executable Line
mpirun  -npernode 1 sss2 -omp 4  msrG12S5_input > nohup.out
touch msrG12S5_input.done