TODO: dissertation reference

This repository holds all the code necessary to run statistical inverse problems to inform surface catalycity model parameters from simulated laser absorption spectroscopy measurements.

Required packages:

1. GCC
2. Autotools
3. MPICH
4. OpenBLAS
5. PETSc
6. HDF5
7. VTK
8. Boost
9. libMesh
10. GRINS
11. GSL
12. Antioch
13. GLPK
14. QUESO

With the packages loaded into PATH and LD_LIBRARY_PATH as needed by the Makefile, one can type make in the src/ directory to build all executables.

1. Generate a noisy data point using generate_data
2. Generate statistical surrogate using generate_surrogate
3. Run optimizer with 1+ noisy data points
4. Run sip with 1+ noisy data points, trim and subsample rawChain, plot histogram to get posterior

• arguments: dimension catalycity_model temperature

• example: mpiexec -np 8 ./generate_data 2 reduced_arr 800

  Generates a single noisy data point for a 2D mesh with the Reduced Arrhenius model for a furnace temperature of 800K

• places data in "./"+dim+"d/"+cat_model+"/noisy_data/noisy_"+temp+".dat"

• arguments: dimension catalycity_model temperature

• example: mpiexec -np 8 ./generate_surrogate 2 constant 800

  Generates a surrogate for a 2D mesh with the Constant model for a furnace temperature of 800K

• places data in "./"+dim+"d/"+cat_model+"/surrogate_data/surr_"+T+".dat"

• arguments: dimension catalycity_model temperatures

• example: mpiexec -np 8 ./optimizer 2 reduced_pwr 600 800

  Solves optimization problem using Nelder-Mead for Reduced Power Law model paramaters using noisy data points for temperatures 600K and 800K

• results printed on-screen

• arguments: dimension catalycity_model temperatures

• example: mpiexec -np 8 ./sip 2 reduced_pwr 600 800 1000

  Solves statistical inverse problem using Metropolis-Hastings for Reduced Power Law model paramaters using noisy data points at temperatures 600K, 800K, and 1000K

• results printed on-screen

• Generates a single noisy data point for temperatures 600,800,1000,1200 K for all 5 catalycity models.
• Data is output to the 2d/[cat_model]/noisy_data/ directories, where it should be copied to 2d/[cat_model]/low_noise_data/ as appropriate based on the run_sip_[cat_model].sh scripts.
• To change flow/laser conditions, edit ./2d/common_input/2d_common.in as needed and then re-run this script.

• Automates all runs of the sip executable to generate inverse problem rawChains that can be trimmed and subsampled using the scripts in the 2d/[cat_model]/analysis directories.
• Expectes data to be in 2d/[cat_model]/low_noise_data/{offset*,xo0*}/ and surrogates in 2d/[cat_model]/surrogates/{offset*,xo)*}/