Latin Hypercube Neutronic Sampling #10

aaswenson · 2018-05-15T17:35:04Z

This PR contains a new module that writes MCNP6 input files. It uses the HomogeneousInput class in mcnp_inputs.py to generate various homogeneous models for sampled reactor configurations.

The main additions are:
neutronic_sweeps.py: generates the inputs
parse_outputs.py: parses the output files for keff results, energy spectra, depletion results, etc

bam241 · 2018-05-15T18:32:06Z

neutronics/neutronic_sweeps.py

+* power
+* enrich
+"""
+from pyDOE import lhs


PEP8: missing blank line

bam241 · 2018-05-15T18:34:53Z

neutronics/neutronic_sweeps.py

+
+from mcnp_inputs import HomogeneousInput, build_pyne_matlib
+
+_dimensions = ['AR', 'core_r', 'cool_r', 'PD', 'power', 'enrich']


doctoring to explain what those variables are.

bam241 · 2018-05-15T19:25:07Z

neutronics/neutronic_sweeps.py

+
+    return sampled, const, dim
+
+


2 lines there

bam241 · 2018-05-15T19:25:50Z

neutronics/neutronic_sweeps.py

+
+    dim = len(sampled)
+
+    return sampled, const, dim


why returning len(sampled) when you actually have sampled?

bam241 · 2018-05-15T19:28:20Z

neutronics/neutronic_sweeps.py

+        cube (ndarray): normalized, N-D latin hypercube
+    """
+
+    np.random.seed(4654562)


why always the same seed ?

If I do the analysis again, I want to be able to generate the same input files.

…tion

gonuke

This could also use some tests.

gonuke · 2018-05-18T00:28:34Z

neutronics/mcnp_inputs.py

@@ -151,7 +151,7 @@ def write_mat_string(self):
        # write mcnp-form string
        self.fuel_string = self.homog_mat.mcnp().strip('\n')

-    def write_input(self):
+    def write_input(self, file_num, header):


Add arguments to docstring....

Not sure what file_num is: is it actually just a number? If so, then not sure about the filename below

gonuke · 2018-05-18T00:29:46Z

neutronics/neutronic_sweeps.py

+
+# Set dimension values. A tuple with range boundaries will sample the dimension
+# evenly using LHS, a float will set the dimension to the constant value
+dims = {'core_r'  : (20, 50),         


Shouldn't _dimensions be set by looking at the keys of this variable to ensure consistency?

gonuke · 2018-05-18T00:35:35Z

neutronics/neutronic_sweeps.py

+            l_limit = dims[dim][0]
+            u_limit = dims[dim][1]
+            # uniform distribution
+            a = u_limit - l_limit


better variable name: width

gonuke · 2018-05-18T00:35:53Z

neutronics/neutronic_sweeps.py

+            u_limit = dims[dim][1]
+            # uniform distribution
+            a = u_limit - l_limit
+            b = l_limit


why not just use l_limit?

gonuke · 2018-05-18T00:37:12Z

neutronics/neutronic_sweeps.py

+            a = u_limit - l_limit
+            b = l_limit
+            # save to ndarray
+            rxt_confs[sample_idx][dim] = b + cube[sample_idx][dim_idx] * a


seems like there is probably a matrix-vector implementation of this?

This won't work as is, but with a little thought on data structures, this whole inner loop can probably be replaced by something like this:
rxtr_confs[sample_idx] = dims[:][0] + cube[sample_idx] * (dims[:][1] - dims[:][0])

gonuke · 2018-05-20T11:13:35Z

neutronics/parse_outputs.py

+            tsteps.append(idx + offset)
+
+    for i, tidx in enumerate(tsteps):
+        step = 0


you can probably just increment tidx in this loop rather than adding step

gonuke · 2018-05-20T11:17:00Z

neutronics/parse_outputs.py

+            ZAID = int(data[1])
+            mass = float(data[2])
+
+            if ZAID in act_inv.keys():


dict.setdefauult() is a handy method to avoid writing these kinds of if statements for dictionaries:

https://www.tutorialspoint.com/python/dictionary_setdefault.htm

https://docs.quantifiedcode.com/python-anti-patterns/correctness/not_using_setdefault_to_initialize_a_dictionary.html

gonuke · 2018-05-20T11:19:22Z

neutronics/parse_outputs.py

+        step.
+        data (ndarray): container to store results
+    """
+    BOL = act_inv[92235][0] / 1000.0


magic number

gonuke · 2018-05-20T11:19:32Z

neutronics/parse_outputs.py

+    BOL = act_inv[92235][0] / 1000.0
+    EOL = act_inv[92235][-1] / 1000.0
+
+    data['BOL_U'] = BOL


why not just assign it directly instead of introducing BOL and EOL

gonuke · 2018-05-20T11:21:31Z

neutronics/parse_outputs.py

+    data['BOL_U'] = BOL
+    data['EOL_U'] = EOL
+
+    delta_rel = abs(EOL - BOL) / data['mass']


why abs? Why not reverse the difference: BOL - EOL, in which case if it's negative, it should be an error...

…or constant and swept parameters

…e some variables for clarity and adjust the tests

gonuke · 2018-06-24T15:38:54Z

neutronics/parse_outputs.py

+        data (ndarray): container to store results
+    """
+    core_r, r, PD, Q = data[['core_r', 'cool_r', 'PD', 'power']]
+    # clad thickness


better to use a clearer variable name than to add a comment to describe your variable name

I used c because it's a common variable used for clad thickness. I can change it though.

aaswenson added 30 commits May 15, 2018 11:59

lhs for parameter space, adjust burnup parameters to shorten calculation

5352233

generate inputs, tar them

b1f3e4e

some htc utiltiles, testing with 40 samples

5fb7b85

--ignore

f995797

A module to parse MCNP depletion outputs

a9ca809

storing results

f6797df

test data

71d51d4

plotting and fitting data

30bba7e

adding core_power

d89deca

don't store strings in memory

b0e8814

initial attempt at multiple regression

786c062

adding mass

50be3ad

calculate mass

a44887c

fix ordering of header parameters

abe77b5

added ability to include constant parameters

ebb7296

data

2e17eb6

ND interpolation of the data

677fa4e

debuggin NN

2f53046

linear fit on the log data, code (rough) processes error

667a118

plot axis labels

1a5e34a

linear regression use sklearn

39216a6

convienience featurees in skit_fit.py

d0101a6

lots of plotting capability, huge commit, kept forgetting to commit...

162d809

trying different regression models

47cfb2a

lots of plotting changes

88839e8

added titles to plotting

e2d25cb

depletion parser

a860836

cleaned up neutronic_sweeps.py nad parse_outputs.py

2f3ac24

more docstrings, comments, and clarifying code changes

0d1e51e

remove depletion results csv

c3f6ffc

docstrings and clarifying comments in parse_outputs.py

22d0077

bam241 reviewed May 15, 2018

View reviewed changes

aaswenson added 2 commits May 16, 2018 10:18

more clarifying comments in neutronics_sweeps.py

4e511cd

set seed for lhs as default optional variable in 'gen_hypercube' func…

4681c38

…tion

gonuke requested changes May 20, 2018

View reviewed changes

aaswenson added 9 commits May 21, 2018 11:01

convert float dimensions to tuples, keep consistency in calculation f…

202f162

…or constant and swept parameters

simplify the parameter range calculation using numpy arrays

73d4418

add reference csv to test output parser, fix energy average

7cd54b2

conversion factors, clarifying comments, removal of magic numbers

390aec9

add pyDOE install to yaml

49a450e

cd into proper directories for testing

c73d6c7

fix variable name

af30af4

pep8 spacing, remove dated docstring from get_sampling_params()

f0c4c66

remove incorrect energy tally parser, simplify the keff parser, renam…

82d4db7

…e some variables for clarity and adjust the tests

gonuke reviewed Jun 24, 2018

View reviewed changes

ypark234 changed the base branch from master to main June 18, 2020 21:24

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Latin Hypercube Neutronic Sampling #10

Latin Hypercube Neutronic Sampling #10

aaswenson commented May 15, 2018

bam241 May 15, 2018

bam241 May 15, 2018

bam241 May 15, 2018

bam241 May 15, 2018

bam241 May 15, 2018

aaswenson May 16, 2018

gonuke left a comment

gonuke May 18, 2018

gonuke May 18, 2018

gonuke May 18, 2018

gonuke May 18, 2018

gonuke May 18, 2018

gonuke May 20, 2018

gonuke May 20, 2018

gonuke May 20, 2018

gonuke May 20, 2018

gonuke May 20, 2018

gonuke Jun 24, 2018

aaswenson Jun 25, 2018


		from mcnp_inputs import HomogeneousInput, build_pyne_matlib

		_dimensions = ['AR', 'core_r', 'cool_r', 'PD', 'power', 'enrich']

Latin Hypercube Neutronic Sampling #10

Are you sure you want to change the base?

Latin Hypercube Neutronic Sampling #10

Conversation

aaswenson commented May 15, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gonuke left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment