Non-Destructive Intelligent Microwave Sensing of Water Content in Sand-Water Mixtures Using S-Parameter Features and...

Area of Science:

• Geotechnical Engineering
• Material Science
• Microwave Engineering

Background:

• Accurate water content estimation is crucial for geotechnical engineering, construction, agriculture, and material assessment.
• Microwave S-parameter measurements offer a non-destructive method for material analysis.

Purpose of the Study:

• To develop and validate a laboratory framework for estimating water content in sand-water mixtures using microwave S-parameters.
• To evaluate the effectiveness of physically interpretable features and various regression models for water content estimation.

Main Methods:

• Acquired 60 microwave S-parameter measurements (S11 and S21) from sand-water mixtures (0-25% water by mass) using a WR-229 waveguide fixture (3.30-4.90 GHz).
• Extracted various descriptors from S-parameter responses, including extrema, slopes, and phase statistics.
• Evaluated six regression models: Ridge, PLS, SVR, GPR, Random Forest, and Gradient Boosting.

Main Results:

• The Partial Least Squares (PLS) regression model achieved an RMSE of 3.56% and R² of 0.826 for estimating water content.
• Leave-one-mixture-level-out cross-validation indicated good generalization performance to unseen mixture levels.
• Magnitude- and phase-derived microwave descriptors, especially transmission features, proved interpretable and repeatable.

Conclusions:

• Microwave S-parameter measurements combined with interpretable features provide a viable framework for water content estimation in granular materials.
• The developed method shows promise for applications in geotechnical engineering and material science.
• Further validation under diverse conditions is necessary for practical deployment.