School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China
Show less
文章历史+
Received
Published
Revised
2025-06-20
2026-06-08
2025-11-28
Online Date
2026-08-07
PDF (8206K)
摘要
纤维金属层合板(fiber-metal laminates,FMLs)因其复合材料的各向异性和层合结构的特征,而展现出优异的可设计性,开展纤维金属层合板的优化设计研究可有效增强其力学性能并实现轻量化目标。为提高纤维金属层合板的抗冲击性能,提出一种基于灰狼优化算法(grey wolf optimizer,GWO)-类别型特征梯度提升(categorical boosting,CatBoost)代理模型纤维金属层合板的铺层角度优化设计方法。采用4因素4水平全因子离散化试验设计,将纤维金属层合板的各铺层角度设为设计变量,同时选取结构吸能和峰值力作为设计目标,根据试验设计方案进行参数化建模获取样本点,建立CatBoost代理模型,构建设计变量与设计目标之间的非线性映射关系,同时采用GWO优化CatBoost的超参数,实现层合板抗冲击性能的高效预测;在此基础上,以层合板的最大吸能和最小峰值力为优化目标,采用遗传算法(genetic algorithm, GA)对层合板的铺层角度进行优化设计。结果表明:GWO-CatBoost代理模型对层合板抗冲击性能的预测具有较高的精准度,可代替复杂耗时的有限元分析(finite element analysis,FEA),有效提高目标方程的计算效率,同时通过显式动力学计算程序ABAQUS/Explicit验证优化效果。在最优的铺层角度方案下,优化后层合板的碰撞峰值力下降18.04%,能量吸收增加了3.77%,其优化设计结果可为工程设计提供参考。
Abstract
Fiber-metal laminates (FMLs) possess significant design flexibility owing to the anisotropic nature of composite materials and their layered structure, making optimization studies essential for enhancing mechanical performance and achieving lightweight objectives. To enhance the impact resistance of FMLs, a novel stacking sequence optimization method was proposed based on a grey wolf optimizer (GWO)-categorical boosting (CatBoost) surrogate model. A four-factor, four-level full factorial design was adopted, using ply angles as design variables with energy absorption and peak force as dual objectives. Parametric modeling generated sample points for establishing a CatBoost surrogate model that captured nonlinear variable-target relationships. The GWO was integrated to optimize CatBoost hyperparameters, enabling efficient laminate impact performance prediction. The stacking sequence of the laminate was then optimized using a genetic algorithm (GA) with dual objectives: maximizing energy absorption and minimizing peak force. The results demonstrate that the GWO-CatBoost surrogate model achieves high-fidelity prediction of laminate impact performance, effectively replacing computationally expensive finite element analysis (FEA) while significantly improving computational efficiency. The optimization outcomes are further validated through explicit dynamic simulations using ABAQUS/Explicit. The optimized laminate demonstrates an 18.04% reduction in peak impact force and a 3.77% improvement in energy absorption under the optimal layup angles, providing a practical reference for engineering design applications.
JU Ming-ming, LIU Fang.
Multi-objective Optimization Design for Impact Resistance of Fiber Metal Laminates Based on GWO-CatBoost Surrogate Model[J].
Science Technology and Engineering, 2026, 26(16): 6793-6803 DOI:10.12404/j.issn.1671-1815.2504600
综上所述,现有研究虽在纤维金属层合板优化设计方面取得显著进展,但仍存在以下关键问题亟待解决:一是传统代理模型(如响应面法、Kriging模型)对类别型设计变量的适应性不足,难以精准捕捉层合板抗冲击性能的复杂非线性特征;二是多目标优化过程中计算效率与模型精度的平衡问题尚未得到系统性解决。为此,提出一种融合灰狼优化算法(grey wolf optimizer,GWO)与类别型特征梯度提升(categorical boosting,CatBoost)代理模型的混合优化框架(GWO-CatBoost),通过以下技术路径实现突破:首先,采用4因素4水平全因子试验设计构建高信息密度的样本空间,解决传统拉丁超立方采样对离散型变量的适配性问题;其次,利用有限元软件ABAQUS建立该层合板结构的数值模型并开展数值模拟研究,通过与文献中的实验数据进行对比验证数值计算方法准确性,引入具有类别特征处理优势的CatBoost算法构建吸能、峰值力多目标代理模型,结合GWO算法实现超参数自适应调优,显著提升模型预测精度;最后,基于遗传算法对铺层角度进行多目标优化,并通过ABAQUS/Explicit显式动力学仿真验证优化方案的有效性。为纤维金属层合板的抗冲击性能优化提供兼具计算效率与工程适用性的新解决方案。
将每个特征变量按30°的增量离散化考虑,每个变量有4个离散水平(0, 30, 60, 90),所有水平组合均为一个样本点,共计44=256组样本点,即采用4因素4水平全因子设计的试验设计;基于ABAQUS有限元分析(finite element analysis,FEA),分别对256组参数化样本实施数值模拟计算,通过后处理模块提取各工况的整体吸能及碰撞峰值力数据,试验设计和数值仿真结果如表5所示。将256组样本数据按8∶2的比例随机划分为训练集和测试集,即从256组样本数据中随机抽取206组作为CatBoost代理模型建立过程中的训练集样本,然后再用GWO算法进行代理模型超参数寻优获取最优超参数进而建立GWO-CatBoost代理模型,最终确定模型的最佳超参数如表6所示,其余50组作为测试集样本。
模型训练完成后,需要在测试集上评估其性能,因此使用50组测试集样本点来检验模型精度,并使用R2和均方根误差(root mean square error,RMSE)来作为评估指标,R2和RMSE的计算方法如式(14)、式(15)所示。其中,R2的取值在[0,1]内,越接近1,表明拟合效果越好,工程中常认为R2>0.9即满足精度要求。RMSE为衡量回归模型误差率的常用公式,其值越小,表明预测值与实际值的偏差越小。
YangZhiwei, ZhangChaoxuan, WangXiaochun, et al. Numerical simulation on damage and failure of missile launcher structure based on difference of composite material layering mode[J]. Science Technology and Engineering, 2025, 25(5): 1920-1926.
XuLiyou, XuZhengzhong, ZhangShuai, et al. Laminate design and multi-level optimization method for front crossbeam of roof carbon fiber composite material in body-in-white[J]. Science Technologyand Engineering, 2025, 25(29): 12451-12465.
[5]
CihanK, ImanM, ZhouJ, et al. High-velocity impact deformation and perforation of fibre metal laminates[J]. Journal of Materials Science, 2018, 53(6): 4209-4228.
ZengYang, ZhouJun, ShenZhiyuan, et al. Optimization design of composite bulkhead structure based on response surface method[J]. Journal of Chongqing University, 2020, 43(6): 82-89.
LiuWeixian, MuXuefeng, ZengGuo. Optimization design of composite honeycomb sandlayer inlet structure[J]. Journal of Nanjing University of Aeronautics and Astronautics, 2018, 50(1): 86-90.
HuChunxing, HouYuliang, TieYing, et al. Influence of different bonding parameters on the strength of CFRP laminates with single lap bonding structure and optimization[J]. Journal of Mechanical Engineering, 2021, 57(8): 154-165.
[16]
EdoreG A, MehdiY. Lay-up optimisation of fibre-metal laminates panels for maximum impact absorption[J]. Journal of Composite Materials, 2020, 54(29): 4591-4609.
[17]
TohidM M, Sayah-BadkhorM, MohammadR, et al. Gas mixture detonation load on polyurea-coated aluminum plates[J]. Thin-Walled Structures, 2020, 155: 106851.
WangZhen, SongKai, ZhuGuohua, et al. Axial energy absorption characteristics of unidirectional carbon-fiber composite cone tubes[J]. Journal of Vibration and Shock, 2018, 37(7): 172-178.
FengZhenyu, SuXuan, ZhaoYanqiang, et al. Optimization design method for energy-absorbing composite structure with probabilistic uncertainty[J]. Journal of Vibration and Shock, 2019, 38(11): 101-109.
[24]
CutoloA, CarotenutoA R, PalumboS, et al. Stacking sequences in composite laminates through design optimization[J]. Meccanica, 2021, 56(6): 1555-1574.
[25]
Ghashochi-BarghH, SadrM H. PSO algorithm for fundamental frequency optimization of fiber metal laminated panels[J]. Structural Engineering and Mechanics, 2013, 47(5): 713-727.
[26]
ZhouY S, ZhengC S. Impact resistance performance and structural optimization of lightweight composite target plate[J]. Polymer Composites, 2024, 45(7): 6482-6497.
[27]
ZhuF, KinneyK, HeW Y, et al. Machine learning guided analysis and rapid design of a 3D-printed bio-inspired structure for energy absorption[J]. Advances in Engineering Software, 2024, 196: 103714.
[28]
LiL Z, HeQ, GuoJ L, et al. Blast response and optimization design of polyurea-coated auxetic honeycomb sandwich panels[J]. Applied Composite Materials, 2023, 30(6): 2047-2070.
[29]
刘芳芳. 玄武岩纤维金属层合板低速冲击性能研究[D]. 长春: 吉林大学, 2020.
[30]
LiuFangfang. Research on low velocity impact performance of basalt fiber metal laminate[D]. Changchun: Jilin University, 2020.
KongXiangshao, YangBao, ZhouHu, et al. Optimal design of ballistic performance of fiber-metal laminates based on the response surface method[J]. Explosion and Shock Waves, 2022, 42(4): 84-96.
[33]
付利荣. 热塑性纤维金属层板抗低速冲击性能分析与优化[D]. 大连: 大连理工大学, 2022.
[34]
FuLirong. Analysis and optimization of low speed impact of thermoplastic fiber metal laminates[D]. Dalian: Dalian University of Technology, 2022.
ZhangQingsong. Optimal design of the ballistic resistance for new titanium alloy composite armor based on SVR surrogate model and NSGA-II algorithm[J/OL]. (2025-01-22)[2025-06-19]. https://doi.org/10.19693/j.issn.1673-3185.04261.
[37]
SitnikovaE, GuanZ W, CantwellW J. The analysis of the ultimate blast failure modes in fibre metal laminates[J]. Composites Science and Technology, 2016, 135: 1-12.
[38]
HashinZ. Failure criteria for unidirectional fiber composites[J]. Journal of Applied Mechanics, 1980, 47(2): 329-334.
[39]
IpsonT W, RechtR F. Ballistic-penetration resistance and its measurement: two impulses reduce the velocity of a projectile passing through a barrier. Minimum perforation velocity is determined by using the ballistic pendulum to measure the impulse due to barrier strength[J]. Experimental Mechanics, 1975, 15(7): 249-257.
[40]
LiudmilaP, GlebG, AleksandrV, et al. CatBoost: unbiased boosting with categorical features[C]// 32nd Conference on Neural Information Processing Systems. New York: Curran Associates Inc., 2018: 9516.
[41]
SeyedaliM, SeyedM M, Andrew L. Grey wolf optimizer[J]. Advances in Engineering Software, 2014, 69: 46-61.
[42]
DebK, PratapA, AgarwalS, et al. A fast and elitist multiobjective genetic algorithm: NSGA-II[J]. IEEE Transactions on Evolutionary Computation, 2002, 6(2): 182-197.