A deep learning-based electrical sensing framework for Structural Health Monitoring of composite structures

This work presents a novel and versatile Structural Health Monitoring (SHM) framework for composite structures, integrating electrical sensing techniques and deep learning. The proposed approach employs a Convolutional Neural Network (CNN) to process electrical potential measurements acquired on the surface of a composite component, enabling the assessment of its structural integrity. By exploiting the CNN’s capability to automatically extract salient features from complex electrical data, the framework allows accurate damage detection, localization and characterization. The methodology is validated through a representative case study involving a Carbon Fibre Reinforced Polymer (CFRP) rear wing. A high-fidelity numerical model is developed to simulate damage scenarios, including translaminar cracks and delaminations at different locations and severity levels, providing the dataset for CNN training, validation and testing. For each damage configuration, the electrical potential is computed at multiple locations, emulating measurements obtained from surface-mounted electrodes. These electrical signatures serve as input features, while the associated damage descriptors are used as target outputs for the CNN. The results demonstrate the accuracy and robustness of the proposed framework in identifying and characterizing damage, highlighting its potential as a computationally efficient, scalable and generalizable strategy for real-time monitoring of complex composite structures.