October 24, 2026
Journal Article
A physics-guided machine learning approach to screening spectral induced polarization measurements
Abstract
Spectral induced polarization (SIP) measurements are increasingly used in repeated monitoring studies, producing datasets that require consistent quality control before interpretation. SIP spectra are information-rich but susceptible to acquisition artifacts, noisy measurements, and nonphysical behavior, making reliable screening essential before spectra are used to infer material or process changes. Existing approaches rely largely on manual expert review or automated classifiers that do not explicitly evaluate physical plausibility. Here, we develop a physics-guided machine-learning workflow for raw SIP quality control that combines CatBoost classification with spectral-shape checks and Kramers-Kronig consistency diagnostics. The workflow was developed from 579 spectra, including 79 labeled spectra used for grouped validation, and assigns spectra to Good, Bad, or Review categories so uncertain cases are retained for inspection rather than forced into binary classifications. In grouped out-of-fold evaluation, both CatBoost-only and physics-guided workflows maintained auto-accept precision of 1.00 and Bad escape rate of 0.00 (i.e., no labeled Bad spectra were accepted as Good). Relative to CatBoost alone, the physics-guided workflow reduced the Good reject rate from 0.60 to 0.45, while introducing a Review pathway for uncertain spectra. These results show that the main value of the workflow is not higher classification accuracy alone, but rather, a more defensible quality-control framework that combines statistical pattern recognition with physical screening. This framework provides a foundation for scalable SIP data validation and for future time-lapse monitoring workflows in which ongoing data streams can be screened consistently before meaningful temporal changes are interpreted.Published: October 24, 2026