Optimizing Unsupervised Clustering of Electrochemical Impedance Spectra via Normalization and Dimensionality Reduction

This dataset supports the study “Optimizing Unsupervised Clustering of Electrochemical Impedance Spectra via Normalization and Dimensionality Reduction” and contains all data and code required to reproduce the reported unsupervised machine-learning analyses of experimental electrochemical impedance spectroscopy (EIS) data.

The repository includes raw EIS spectra of welded 316L stainless steel measured in aerated 0.1 M NaCl, representing different surface conditions (base metal vs. heat-affected zone, mechanically cleaned, as-welded, and citric-acid-passivated states), together with abraded and nitric-acid-passivated reference samples. Spectra are provided as frequency-resolved Bode data (log|Z| and phase).

The analysis workflows implement four preprocessing strategies: (i) raw (no normalization), (ii) per-block min–max normalization, (iii) per-sample min–max normalization, and (iv) per-column autoscaling. These datasets serve as input for dimensionality reduction using principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), and combined PCA + t-SNE pipelines.

The repository further includes Orange (.ows) workflow files, custom Python scripts, clustering results, linear projection plots, silhouette analyses, hierarchical dendrograms, and bootstrap resampling outputs used to assess clustering quality and stability. Validation metrics (Silhouette score, Davies–Bouldin index, Calinski–Harabasz index, ARI, NMI, and Purity) are provided for all evaluated pipelines.

All files are organized into structured folders (Input Data, Code, Graphs, Results) to facilitate transparent reuse and reproducibility. The dataset is intended for researchers working on electrochemical impedance spectroscopy, corrosion science, and machine-learning-based spectral analysis, and may serve as a benchmark for unsupervised analysis of experimental EIS data.