Skip to content

four2five/simform

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

simform: MapReduce based Simulation Informatics

Written by Austin Benson, Paul Constantine, David F. Gleich, and Yangyang Hou

Functionality

This code is designed to work with a database of exodus files produced by the ARIA thermal code in the Sierra Mechanics toolbox. For such a database, it will:

  • convert the data into a format to make it easier to process in Hadoop
  • compute the SVD of the database
  • interpolate the solution at any set of new points using a linear interpolation routine
  • interpolate the solution at any set of new points using a non-linear model reduction based interpolation routine that includes a estimate of the error
  • convert all the MapReduce based predictions to exodus files.

Synopsis

Here are some commands I ran to do a quick analysis on the EC2 cluster:

Initialization

$ make dir=hdfs://nebula/data/exodus-runs

$ make setup_database name=runs variable=TEMP dir=hdfs://ec2-107-22-80-153.compute-1.amazonaws.com:8020/user/temp/simform/
$ make -f runs preprocess
$ make -f runs convert timestepfile=timesteps.txt

In this case, we had to normalize time-steps across the different files as the default step-length is variable.

Simple interpolation

$ make -f runs predict design=design_points.txt points=new_points.txt

and then dump out exodus files

$ make -f runs seq2exodus  numExodusfiles=10 OutputName=output/thermal_maze

SVD based Model Reduction

$ make -f runs seq2mseq
$ make -f runs model numExodusfiles=6
$ make -f runs interpsvd design=design_points.txt points=new_points.txt

Setup

mrsimform requires quite a few libraries as it combines the functionality of many different pieces of code. Please see INSTALL.md for help on setup. The code does not actually require any pieces to be installed on the system. But this makes it much easier. We've provided a list of commands to setup a demostration cluster on Amazon's EC2 system and process a small collection of files. See the demo directory.

Papers

These codes are based on a few active research projects from the authors.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%