[Paper Review] Crack-Net: Prediction of Crack Propagation in Composites

Hao Xu, Wei Fan+4 more|arXiv (Cornell University)|Sep 24, 2023

Numerical methods in engineeringEngineering3 citations

TL;DR

Crack-Net is a deep learning framework that predicts crack propagation and stress-strain responses in composite materials by learning the relationship between microstructure, stress fields, and fracture evolution. Trained on high-fidelity phase field simulation data, it achieves accurate, fast predictions of long-term crack growth—even in complex microstructures like binary co-continuous composites—using transfer learning to enhance generalization across material types.

ABSTRACT

Computational solid mechanics has become an indispensable approach in engineering, and numerical investigation of fracture in composites is essential as composites are widely used in structural applications. Crack evolution in composites is the bridge to elucidate the relationship between the microstructure and fracture performance, but crack-based finite element methods are computationally expensive and time-consuming, limiting their application in computation-intensive scenarios. Here we propose a deep learning framework called Crack-Net, which incorporates the relationship between crack evolution and stress response to predict the fracture process in composites. Trained on a high-precision fracture development dataset generated using the phase field method, Crack-Net demonstrates a remarkable capability to accurately forecast the long-term evolution of crack growth patterns and the stress-strain curve for a given composite design. The Crack-Net captures the essential principle of crack growth, which enables it to handle more complex microstructures such as binary co-continuous structures. Moreover, transfer learning is adopted to further improve the generalization ability of Crack-Net for composite materials with reinforcements of different strengths. The proposed Crack-Net holds great promise for practical applications in engineering and materials science, in which accurate and efficient fracture prediction is crucial for optimizing material performance and microstructural design.

Motivation & Objective

• To address the computational cost of traditional finite element methods in simulating crack propagation in composites.
• To develop a data-driven framework that captures the physics of crack evolution from stress fields and microstructure.
• To enable efficient prediction of long-term crack growth patterns and stress-strain curves for diverse composite designs.
• To improve generalization across different reinforcement strengths using transfer learning.
• To support practical engineering design by accelerating fracture performance prediction in structural composites.

Proposed method

• Crack-Net employs a convolutional neural network (CNN) architecture to map input microstructures and applied stress fields to predicted crack paths and evolution.
• The model is trained on a large-scale dataset of fracture simulations generated via the phase field method, ensuring high accuracy in crack morphology and energy dissipation.
• Input features include the initial microstructure and stress distribution, while outputs are spatial crack evolution maps and corresponding stress-strain curves.
• The framework incorporates attention mechanisms to emphasize critical stress concentration regions influencing crack nucleation and propagation.
• Transfer learning is applied to fine-tune the model on new composite systems with varying reinforcement strengths, improving zero-shot and few-shot generalization.
• The model is validated against phase field simulations and tested on complex microstructures, including binary co-continuous morphologies.

Experimental results

Research questions

• RQ1 Can a deep learning model accurately predict long-term crack propagation patterns in composite materials from initial microstructure and stress fields?
• RQ2 How well does the model generalize to composite microstructures with different reinforcement strengths?
• RQ3 Can the model capture complex crack morphologies, such as those in binary co-continuous composites, with high fidelity?
• RQ4 To what extent does transfer learning improve prediction accuracy for unseen composite systems?
• RQ5 Can the model predict the full stress-strain response alongside crack evolution, enabling performance assessment?

Key findings

• Crack-Net achieves high accuracy in predicting crack paths and evolution patterns, closely matching results from computationally expensive phase field simulations.
• The model successfully predicts crack propagation in complex binary co-continuous microstructures, demonstrating robustness to morphological complexity.
• Transfer learning significantly improves generalization, enabling accurate predictions for composite materials with reinforcement strengths not seen during pre-training.
• The model reduces inference time by several orders of magnitude compared to traditional finite element methods, enabling real-time or near-real-time fracture analysis.
• The predicted stress-strain curves show strong agreement with reference simulations, validating the model’s ability to capture mechanical response alongside damage evolution.
• Attention maps highlight critical stress concentration zones, confirming the model learns physically meaningful crack initiation mechanisms.