Importance Sampling and Feature Fusion Paradigm-Boosted Multi-Modal Convolutional Neural Networks

Deployment in Composite Curing Process Monitored by Electro-Mechanical Impedance

Journal Article (2025)
Author(s)

Zeyuan Gao (Civil Aviation Flight University of China)

Xin Zhao (Civil Aviation Flight University of China)

Meng Li (Civil Aviation Flight University of China)

Z. Han (TU Delft - Group Turteltaub)

Jianjian Zhu (Civil Aviation Flight University of China)

Research Group
Group Turteltaub
DOI related publication
https://doi.org/10.1109/ACCESS.2025.3551508
More Info
expand_more
Publication Year
2025
Language
English
Research Group
Group Turteltaub
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

The increasing application of composite materials in various industrial sectors is driven by their lightweight nature, high strength-to-stiffness ratio, and corrosion resistance. Effective monitoring of the curing process is crucial for ensuring quality and performance. Electro-Mechanical Impedance (EMI) offers promising, non-destructive, real-time monitoring, but the complexity of EMI signals poses challenges. Convolutional Neural Networks (CNNs) have the potential to enhance EMI-based monitoring accuracy. However, training CNNs on multi-modal EMI signals requires addressing data heterogeneity, class imbalance, and computational complexity at present. This study develops the Importance Sampling Algorithm-optimized Multi-Modal CNNs (ISA-MM-CNNs) paradigm for EMI-based evaluation of composite curing processes. By prioritizing informative samples and capturing complementary information from diverse EMI signal modalities, we aim to improve the robustness and efficiency of CNNs in evaluating curing degrees. This study outlines EMI monitoring challenges, details the ISA-MM-CNNs paradigm, and discusses data preprocessing, network architecture, and training optimization. Experimental results demonstrate the superiority of the developed ISA-MM-CNNs and suggest further studies for the curing monitoring of composites.