Impedance-Based Estimation of Process Parameters in Electrolytic Systems via Circuit-Embedded Neural Networks (CENN) - SOURCE CODES

The Modeling_ML_Scripts folder contains Jupyter notebooks and Python code used for the impedance-based estimation of electrolyte concentration and temperature in aqueous H₂SO₄ with nanoporous gold electrodes. It covers the full workflow from raw EIS data to inverse estimation. Scripts handle data extraction and preprocessing (cutting spectra to selected frequency bands, fitting, and interpolation), physics-aware batch fitting of the ZARC + transmission-line equivalent circuit across the (C,T) domain, and extraction of solution resistance and conductivity with degeneracy analysis. The core modeling includes a Circuit-Embedded Neural Network (CENN) forward model that predicts impedance from concentration (introduced for the first time by the author), temperature, and frequency, and an inverse pipeline that estimates C and T from measured spectra using gradient-based optimization. Sensitivity analysis scripts compute Jacobian-based frequency rankings and band-wise contributions of Z′ and −Z″ to concentration and temperature. Classical machine learning benchmarks (Ridge, SVR, GPR, MLP, Random Forest, etc.) are provided for comparison with the CENN. The folder also includes a requirements file for dependencies and a README with installation instructions, script descriptions, and usage guidance for reproducibility.