Application of Machine Learning Methods for Moisture Estimation in Masonry Materials Using a Single-Antenna Microwave Technique

Moisture in building materials, particularly in masonry structures, can negatively affect the durability, serviceability, and safety of buildings. Therefore, it is important to develop non-destructive diagnostic methods that enable reliable assessment of moisture content. Microwave techniques provide a method for determining a material’s moisture content indirectly from changes in its electromagnetic properties, particularly the apparent permittivity. The aim of this article is to present the application of machine learning methods for estimating the moisture content of masonry materials based on laboratory microwave measurements. The study uses a measurement setup with a single antenna operating simultaneously as both transmitting and receiving antenna, which increases the practical applicability of the method in in-situ diagnostics but also makes the interpretation of the measured signal more challenging. Predictive models were developed using the scikit-learn library and dedicated software designed to automatically select the best-performing model and its parameters. The obtained results make it possible to assess the potential of machine learning as a tool supporting the interpretation of microwave data and improving the accuracy of non-destructive moisture diagnostics in masonry structures.