Intelligent monitoring of impact damage within concrete through deep learning-empowered electromechanical impedance technique

As fundamental elements in civil infrastructures, concrete structures may experience impact loads during service life, which poses threat to the structural integrity and serviceability. Accurate detection and assessment of internal damage are critical to ensuring post-impact safety and guiding reinforcement strategies. The electromechanical impedance (EMI) technique has proven to be a reliable non-destructive approach for detecting concrete damage. However, traditional EMI approaches rely on manual feature extraction and statistical analysis, hindering real-time and intelligent applications. To address this limitation, this paper developed a fused deep learning framework named KoCG-Net, which integrates convolutional neural networks (CNNs), gated recurrent units (GRUs), and Kolmogorov-Arnold networks (KAN) to automate the EMI-based damage detection. KoCG-Net directly processed raw conductance signals from repetitive drop weight impact tests, achieving accurate prediction of impact damage. The results demonstrated its superior performance, with R² values of 0.9937 for C30 dataset and 0.9985 for C50 dataset. Moreover, the framework outperformed five benchmark models in prediction accuracy, noise immunity, and efficiency under limited training data, manifesting its substantial potentials for real-time and intelligent monitoring of impact damage within concrete.