Non-intrusive microwave technique for direct detection of concrete compressive strength monitoring by multivariate modeling

This paper investigates using microwaves from 1.5 to 14 GHz for monitoring concrete compressive strength in real-time to implement a non-intrusive sensor. The approach is based on acquiring reflected microwave spectra by the concrete irradiated through a rectangular cavity antenna connected to a Vector Network Analyzer (VNA) and investigating their relationship with strength. Since the information is hidden in the spectrum, we used the multivariate statistical analysis technique to create both classification and regression models. The acquired Voltage Standing Wave Ratio (VSWR) spectra were statistically inferred with compressive strength from destructive tests. The experimentation was performed on 16 concrete samples, lasted 28 days, and the acquired data consisted in 409 spectra where the reflected spectra are correlated with the compressive strength during curing time. The inference processes show high predictability, with an F1 score from 0.91 to 0.99 for classifi-cation and a cross-validation value of the parameter R2 of 0.997.