Na Li , Dawei Sun , Zhiwei Xing , Wanxin Huang , Qiuhan Wang
{"title":"Hygrothermal-vibration coupled aging prediction of CFRP laminates via multiscale modeling and deep learning","authors":"Na Li , Dawei Sun , Zhiwei Xing , Wanxin Huang , Qiuhan Wang","doi":"10.1016/j.coco.2025.102584","DOIUrl":null,"url":null,"abstract":"<div><div>Carbon fiber reinforced polymer (CFRP) laminates applied in aircraft engine nacelles suffer complex coupled aging effects of hygrothermal and vibration loads, while existing studies rarely consider such multi-physics coupling conditions. This study developed a novel experimental platform capable of simultaneously applying hygrothermal and vibration loads to CFRP laminates, combined with multiscale numerical simulations and deep learning approaches for accurate aging prediction. Experimental tests yield bending performance data under limited aging durations due to time and cost constraints, resulting in coarse fitting curves and insufficient accuracy for performance prediction at arbitrary aging times. Concurrently, continuous monitoring of electrical resistance during aging revealed significant conductivity degradation, providing complementary insight into progressive damage potentially linked to fiber integrity, an aspect often overshadowed by matrix degradation studies. To address these data limitations and capture internal mechanisms, a multiscale finite element method (FEM) model incorporating temperature and humidity-dependent material properties is established to obtain internal stress–strain states and dynamic microscopic damage processes inaccessible through direct experiments. Time-series Generative Adversarial Networks (TimeGAN) augment the experimental dataset, while an improved Gated Recurrent Unit (GRU) -enhanced graph neural network (GNN) captures the temporal evolutions of internal stress–strain and environmental parameters, enabling precise predictions of residual mechanical properties. Validation experiments confirm superior prediction accuracy compared to traditional fitting methods, providing robust guidance for aerospace composite reliability assessment.</div></div>","PeriodicalId":10533,"journal":{"name":"Composites Communications","volume":"59 ","pages":"Article 102584"},"PeriodicalIF":7.7000,"publicationDate":"2025-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Composites Communications","FirstCategoryId":"88","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2452213925003377","RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MATERIALS SCIENCE, COMPOSITES","Score":null,"Total":0}
引用次数: 0
Abstract
Carbon fiber reinforced polymer (CFRP) laminates applied in aircraft engine nacelles suffer complex coupled aging effects of hygrothermal and vibration loads, while existing studies rarely consider such multi-physics coupling conditions. This study developed a novel experimental platform capable of simultaneously applying hygrothermal and vibration loads to CFRP laminates, combined with multiscale numerical simulations and deep learning approaches for accurate aging prediction. Experimental tests yield bending performance data under limited aging durations due to time and cost constraints, resulting in coarse fitting curves and insufficient accuracy for performance prediction at arbitrary aging times. Concurrently, continuous monitoring of electrical resistance during aging revealed significant conductivity degradation, providing complementary insight into progressive damage potentially linked to fiber integrity, an aspect often overshadowed by matrix degradation studies. To address these data limitations and capture internal mechanisms, a multiscale finite element method (FEM) model incorporating temperature and humidity-dependent material properties is established to obtain internal stress–strain states and dynamic microscopic damage processes inaccessible through direct experiments. Time-series Generative Adversarial Networks (TimeGAN) augment the experimental dataset, while an improved Gated Recurrent Unit (GRU) -enhanced graph neural network (GNN) captures the temporal evolutions of internal stress–strain and environmental parameters, enabling precise predictions of residual mechanical properties. Validation experiments confirm superior prediction accuracy compared to traditional fitting methods, providing robust guidance for aerospace composite reliability assessment.
期刊介绍:
Composites Communications (Compos. Commun.) is a peer-reviewed journal publishing short communications and letters on the latest advances in composites science and technology. With a rapid review and publication process, its goal is to disseminate new knowledge promptly within the composites community. The journal welcomes manuscripts presenting creative concepts and new findings in design, state-of-the-art approaches in processing, synthesis, characterization, and mechanics modeling. In addition to traditional fiber-/particulate-reinforced engineering composites, it encourages submissions on composites with exceptional physical, mechanical, and fracture properties, as well as those with unique functions and significant application potential. This includes biomimetic and bio-inspired composites for biomedical applications, functional nano-composites for thermal management and energy applications, and composites designed for extreme service environments.