Intelligent Early-Corrosion Detection System Based on Backscattering Sensors

Since water and oxygen in the pipeline coatings are primary causes of metallic corrosion, predicting potential moisture exposure is crucial for early detection of metallic corrosion. An intelligent detection system utilizing back-scattering microwave sensors is proposed and demonstrated, which is applied to early metallic corrosion detection. The back-scattering sensors comprise nine double U-shaped resonant units, generating nine resonant peaks within the frequency range of 3.5 to 7.5 GHz. These resonant frequencies shifts corresponding to the water ingress conditions of the metal surface coating. The minimum detectable water ingress volume of the sensor is 1 mm<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup>. Radar cross section (RCS) data under various water ingress conditions are collected through simulations and experiments. Borderline-SMOTE oversampling is adopted as a preprocessing step to address the dataset class imbalance problem. After resonant frequencies are extracted as features, multiple machine learning classification algorithms have been applied and compared for modeling and training, with random forest showing superior performance. The classification accuracy of the simulation and experimental datasets reaches up to 98.6% and 95.8%, respectively. The proposed system can effectively identify whether water ingress occurs inside the protective layer to be tested, which facilitates the low-cost monitoring of the healthy state of the coating of metals and enabling timely intervention.