基于人工神经网络海因环氧树脂体系流变性的比较研究

doi:10.19936/j.cnki.2096-8000.20250728.012

摘要/Abstract

摘要： 化学流变学分析在热固性树脂基复合材料加工成型中具有广泛应用,然而传统化学流变模型对加工黏度的预测常使用Dual-Arrhenius流变模型,该方法在具有多场耦合效应的热固性树脂体系中预测精度受限。为突破这一瓶颈,引入了人工智能的方法。采用DSC法分析了海因环氧树脂/马来酸酐体系的反应性,在65~85 ℃等温固化模式下进行了黏度数据采集,采用BP人工神经网络模型(BP-ANN模型)与Dual-Arrhenius流变模型(DA模型)对海因环氧树脂/马来酸酐体系的流变性能进行对比分析研究。结果表明,相较于DA模型,BP-ANN模型的均方误差降幅达26.0%,平均绝对百分比误差锐减65.0%,均方根误差降低13.0%,同时决定系数提升0.25%。BP-ANN模型相较于DA模型在预测精度上显著提升,这对热固性树脂基复合材料成型过程中的工艺选择和参数优化提供了一定的支撑。

关键词: 海因环氧树脂, Dual-Arrhenius流变模型, BP人工神经网络模型, 化学流变学, 黏度, 复合材料

Abstract: Chemorheological analysis has been widely applied in the processing of thermoset matrix composites. However, traditional chemorheological models for predicting processing viscosity, such as the Dual-Arrhenius (DA) rheological model, exhibit limited accuracy in thermoset resin systems with multi-field coupling effects. To address this limitation, an artificial intelligence (AI)-based approach was introduced. The reactivity of the hydantoin epoxy resin/maleic anhydride (HY/MAD) system was investigated using differential scanning calorimetry (DSC), and viscosity data were collected under isothermal curing conditions at 65~85 ℃. A backpropagation artificial neural network (BP-ANN) model was developed and systematically compared with the DA model to analyze the rheological behavior of the HY/MAD system. The results demonstrate that the BP-ANN model significantly outperforms the DA model in predictive accuracy: the mean square error is reduced by 26.0%, the mean absolute percentage error decreases sharply by 65.0%, the root mean square error is lowered by 13.0%, and the coefficient of determination improves by 0.25%. This marked enhancement in prediction precision provides critical support for optimizing process parameters and material design in thermoset matrix composite manufacturing, particularly in scenarios involving complex multi-physics coupling.

Key words: hydantoin epoxy resin, Dual-Arrhenius rheological model, backpropagation artificial neural network, chemorheological, viscosity, composites

中图分类号:

TB332

雷世裕, 李玲, 曹伟, 王艺轩, 董夏瑞. 基于人工神经网络海因环氧树脂体系流变性的比较研究[J]. 复合材料科学与工程, 2025, 0(7): 99-107.

LEI Shiyu, LI Ling, CAO Wei, WANG Yixuan, DONG Xiarui. Comparative study of chemorheological properties in hydantoin epoxy resin system using artificial neural networks[J]. COMPOSITES SCIENCE AND ENGINEERING, 2025, 0(7): 99-107.

参考文献

[1] 文友谊, 刘立朋, 叶宏军, 等. BMI树脂化学流变特性及化学流变模型研究[J]. 工程塑料应用, 2015, 43(5): 101-104.
[2] 赵卫生, 孙超明, 王文晶, 等. RTM用乙烯基酯树脂化学流变性研究 [J]. 玻璃钢/复合材料, 2014(8): 97-100.
[3] 张旭锋, 黎迪辉, 齐僖, 等. 基于松香酸酐的生物质环氧树脂体系流变特性[J]. 航空材料学报, 2021, 41(2): 98-104.
[4] 高堂铃, 付刚, 王先杰, 等. RTM用低黏度树脂体系的工艺窗口预测[J]. 哈尔滨工业大学学报, 2021, 53(2): 162-167.
[5] WU Z, WANG H, HE C, et al. The application of physics-informed machine learning in multiphysics modeling in chemical engineering[J]. Industrial & Engineering Chemistry Research, 2023, 62(44): 18178-18204.
[6] ZHOU Z, WANG L, WANG J, et al. Machine learning with neural networks to enhance selectivity of nonenzymatic electrochemical biosensors in multianalyte mixtures[J]. ACS Applied Materials & Interfaces, 2022, 14(47): 52684-52690.
[7] WANG T, WU B, HE Y, et al. AI prediction of H₂ production in a biomass supercritical water gasification fluidized bed under pulsating inlets[J]. Industrial & Engineering Chemistry Research, 2023, 62(42): 17182-17193.
[8] SHAH H, ILYAS A, MADRY A. Decomposing and editing predictions by modeling model computation[EB/OL]. (2024-04-17) [2025-04-17]. https://doi.org/10.48550/arXiv.2404.11534.
[9] 罗玲, 田智立, 张涛, 等. 基于人工神经网络的热固性树脂基复合材料固化变形预测研究综述[J]. 复合材料科学与工程, 2022(11): 120-127.
[10] HU Y, ZHAO W, WANG L, et al. Machine-learning-assisted design of highly tough thermosetting polymers[J]. ACS Applied Materials & Interfaces, 2022, 14(49): 55004-55016.
[11] 牛芳旭, 孙超明, 贺靖, 等. 共固化蜂窝夹层结构长梁弯曲性能预测[J]. 复合材料科学与工程, 2025(2): 34-39, 128.
[12] 闫珊, 付天宇, 许家忠, 等. 基于改进遗传算法的纤维张力模糊控制研究[J]. 复合材料科学与工程, 2025(2): 54-61, 144.
[13] SHARAN A, MITRA M. Prediction of static strength properties of carbon fiber-reinforced composite using artificial neural network[J]. Modelling and Simulation in Materials Science and Engineering, 2022, 30(7): 075001.
[14] 刘敬福, 叶建军, 周祥春, 等. 喷射成形TiC_p/ZA35复合材料热挤压工艺的ANN优化和组织研究[J]. 航空材料学报, 2023, 43(2): 59-65.
[15] 李骁, 夏佳佳, 张向奎. 基于人工神经网络的短纤维增强复合材料设计[J]. 计算机辅助工程, 2024, 33(2): 17-23.
[16] XIN Y Q, LI L, FAN L Y. Prediction and analysis on chemorheology of MeHHPA/hydantoin epoxy resin[J]. Key Engineering Materials, 2017, 727: 497-502.
[17] 李玲, 信雅全, 秦旭锋, 等. 海因环氧树脂的合成及其与六氢苯酐的固化行为[J]. 表面技术, 2018, 47(1): 230-235.
[18] XING W, LI L, WANG Y, et al. Recyclable and shape-memory hydantoin epoxy resins based on dynamic ester-exchanged bonds[J]. Polymer Engineering & Science, 2023, 64(2): 506-517.
[19] BLANCO J, LINARES M, LÓPEZ GRANADOS M, et al. Integrated environmental and exergoeconomic analysis of biomass-derived maleic anhydride[J]. Advanced Sustainable Systems, 2022, 6(9):2200121.
[20] 梁志勇, 段跃新, 林云, 等. EPON862环氧树脂体系化学流变特性研究[J]. 复合材料学报, 2001, 18(1): 16-19.
[21] 吴海, 肖加余, 邢素丽, 等. 耐高温缩水甘油胺型环氧树脂的化学流变特性[J]. 复合材料学报, 2016, 33(4): 741-748.
[22] KIUNA N, LAWRENCE C J, FONTANA Q P V, et al. A model for resin viscosity during cure in the resin transfer moulding process[J]. Composites Part A: Applied Science and Manufacturing, 2002, 33(11): 1497-1503.
[23] 于志省, 李应成, 王宇遥, 等. 人工神经网络在材料开发中的应用研究进展[J]. 工程塑料应用, 2023, 51(2): 158-164.
[24] CAO L. A new age of AI: features and futures[J]. IEEE Intelligent Systems, 2022, 37(1): 25-37.
[25] WANG M, XU J, REN W, et al. Laser electrochemical deposition hybrid preparation of an oil-water separation mesh with controllable pore diameter based on a BP neural network[J]. Langmuir, 2023, 39(21): 7281-7293.
[26] WANG Y, WANG W, CHEN Y. Carnivorous plant algorithm and BP to predict optimum bonding strength of heat-treated woods[J]. Forests, 2023, 14(1): 51.