Detection of Iron Corrosion in Concrete Using 1D Convolution

Iron corrosion in concrete poses significant risks to infrastructure durability and integrity. This study presents a novel method for detecting iron corrosion using a one-dimensional convolutional neural network (1D-CNN). Concrete samples (30×20x10 cm, K125 grade) were immersed in 3% salt water and electrified with a 12V, adapter to accelerate corrosion. Data was collected with a calibrated Lite VNA 64, generating S1P and S2P files from corroded and non-corroded samples, which were converted to npy format for 1D-CNN model training. The 1D-CNN used convolutional and pooling layers to extract features and detect corrosion, achieving high accuracy, as shown by F1 score, precision, recall, and accuracy metrics.