Machine Learning based Analysis of LIBS and EIS Data to estimate the Moisture Content of Soil in Grounding Systems

The dielectric properties of soils are significantly influenced by moisture content, which directly affects the effectiveness of the grounding system. This study presents a novel approach that integrates Laser-Induced Breakdown Spectroscopy (LIBS) and Electrochemical Impedance Spectroscopy (EIS) with machine learning to accurately estimate soil moisture. Experiments were conducted on mixtures of bentonite and sand in various proportions and with a wide range of moisture contents. From EIS data, bulk resistance and constant-phase-element parameters were extracted as indicators of moisture-driven dielectric response, while LIBS provided elemental emission intensities and plasma temperature. A Light Gradient Boosting Machine (LightGBM) regression model trained on the combined feature set provided excellent prediction accuracy (R² ≈ 0.975). The Shapley Additive exPlanations (SHAP) analysis enhanced the model's interpretability by identifying key predictors, such as hydrogen emission, the H/Na ratio, and soil resistance. The physical understanding of moisture-dependent soil behavior has improved due to increased interpretability. The proposed LIBS–EIS approach enables fast, non-destructive, and interpretable monitoring of soil dielectric behavior.