A hybrid electromechanical phase-field and deep learning framework for predicting fracture in dielectric nanocomposites

The accurate and efficient prediction of crack propagation in dielectric materials is a critical challenge in structural health monitoring and the design of smart systems. This work presents a hybrid modeling framework that combines an electromechanical phase-field fracture model with deep learning-based surrogate modeling to predict fracture evolution in dielectric nanocomposite plates. The underlying finite element simulations capture the coupling between mechanical deformation and electrical field perturbations caused by cracks, using a variational phase-field formulation. High-fidelity simulation outputs–namely, phase-field damage variables and electric potential fields–are utilized to train convolutional neural networks (CNNs) with ResNet-U-Net architectures. Crucially, the framework is designed to predict the final crack path directly from the geometric configuration (random defect patterns). We systematically compare the effectiveness of using either phase-field variables or electric potential fields as the primary physical signatures to guide the training process. The results reveal that models informed by electric potential fields offer superior segmentation accuracy, faster convergence, and enhanced generalization, owing to the smoother gradient distribution and global spatial coverage of the electrical response. Ultimately, the trained surrogate model enables the instantaneous, geometry-driven prediction of crack paths, bypassing the need for computationally intensive field calculations during inference. This demonstrates that leveraging electrical physics as a “training guide” significantly improves the reliability of real-time fracture assessment in smart materials. • Thermodynamically consistent phase-field model captures coupled stiffness degradation and permittivity loss in dielectric nanocomposites. • Large-scale high-fidelity dataset of 10,000 electromechanical fracture simulations generated for surrogate model training. • ResNet-U-Net framework developed for pixel-wise crack segmentation using both mechanical and electrical signatures. • Electric potential fields found to outperform phase-field data as training inputs, offering higher accuracy and faster convergence. • Hybrid physics–deep learning approach enables real-time fracture prediction and structural health monitoring for smart composites.