Convolutional Neural Network‐Based Constitutive Modeling and Parameter Identification for Aluminum/Epoxy Resin Composites

ABSTRACT The novel reactive material aluminum/epoxy resin (Al/Ep) demonstrates significant potential for applications in damage assessment and blast mitigation research owing to its exceptional impact‐induced energy release characteristics. To systematically investigate the mechanical properties of Al/Ep composites, this study develops a comprehensive framework involving constitutive modeling, numerical simulation, and parameter identification. First, Al/Ep specimens with controlled variations in Al volume fractions are fabricated and subjected to rigorous quasi‐static and dynamic mechanical testing. Second, leveraging the Helmholtz free energy framework and Boyce–Parks–Argon (BPA) constitutive theory, a sophisticated constitutive model that accurately captures the complete elastoplastic deformation behavior of Al/Ep materials is formulated. A finite element numerical model with the material user subroutine is established. Third, a robust convolutional neural network‐based analysis methodology is developed to establish a high‐fidelity mapping between stress–strain responses and constitutive parameters, effectively addressing the challenge of multi‐parameter identification for heterogeneous material constitutive models. The experiment confirms the accuracy and reliability of the proposed research framework. This comprehensive framework provides an efficient and precise means for evaluating both static mechanical properties and dynamic response characteristics of Al/Ep composites with varying compositions, thereby significantly advancing the development and performance optimization of epoxy‐based composite materials.