PyCatRobin

Python module to analyze time-on-stream Catalyst testing results from Round Robin test. Catalyst durability is an essential component to secure to realize industrial catalysts. A popular form of data that characterizes the durability is time-series data. When it comes to increasing dataset size by combining multiple data sources like reactors, we face challenges such as systematic errors between reactors and varying experimental conditions. PyCatRobin is developed to address these challenges by providing functionalities to extract features from time-series data, quantify uncertainties of them due to heterogeneity in reactors/conditions, and visualize the results effectively.

Requirements

• All specified in setup.py

Getting started

1. Make a virtual environment (e.g., when using conda):

conda create -n pycatrobin python=3.12
conda activate pycatrobin

2. Installation

• choice1) Directly install using pip

  pip install git+https://github.com/dongjae-shin/PyCatRobin.git

• choice2) Clone repository & install using pip

  git clone https://github.com/dongjae-shin/PyCatRobin.git
  cd PyCatRobin
  pip install .

3. Run example codes (under development)

• Example python codes to use pycatrobin are in examples/ directory.
• In the examples/, run python scripts as follows:

  python ./Welchs_t_test_Fig_3.py
  python ./fANOVA.py

• Currently, Welch's t-test and fANOVA codes are separate scripts from pycatrobin. They will be incorporated into the main package in the near future.
• Run also Jupyter Notebooks such as SNR_heatmap_Fig_3.ipynb, Feature_impact_Fig_S30.ipynb, and NV_heatmaps_Figs_S28-29.ipynb as is in the following example:

  %matplotlib inline
  
  import pycatrobin.data.extract as ex
  import pycatrobin.analysis.data_analysis as da
  
  (...)
  
  # Specify the order of methods and properties to plot
  methods=[
      'AUC',
      'final value',
      'initial value',
      'final slope',
      'initial slope',
      'overall slope'
      ]
  
  properties=[
      'CH4 Net Production Rate (mol/molRh/s)',
      'CO Net Production Rate (mol/molRh/s)',
      'CO2 Conversion (%)',
      'Selectivity to CO (%)'
      ]
  
  # Plot heatmap of SNR values
  analysis.plot_heatmap(
      methods=methods,
      properties=properties,
      which_to_plot='snr', # std_dev, std_dev_mean_normalized
      snr_type='mu_sigma', # 'std_dev', 'range'
      cmap='Reds', # 'Blues'
      vmax=5.3,
      # vmin=0.0,
  )

  

Related publication

• Quantifying Experimental Uncertainty in Catalyst Deactivation: Round-Robin Testing and Implications for Machine-Learned Prediction*, S. Bac, D. Shin, S. Hong, J. Heinlein, A. Khan, G. Barber, Z. Chen, M. M. Albrechtsen, C. Tassone*, R. M. Rioux*, M. Cargnello*, S. R. Bare*, K. Winther*, P. Christopher*, A. S. Hoffman*, submitted (2025).

Acknowledgement

• Original codes for t-test and fANOVA analyses were written by Dr. Selin Bac (UCSB) and Michael Albrechtsen (DTU), respectively.