Yang Xiaobo, Zou Xingyu, Zhang Jingyu, Guo Ruiqing, Xiang He, Zhan Lihua, Wu Xintong
College of Mechanical and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China.
Aerospace Research Institute of Materials & Technology, Beijing 100079, China.
Polymers (Basel). 2025 Jun 26;17(13):1769. doi: 10.3390/polym17131769.
The epoxy-bonded joint between carbon-fiber-reinforced bismaleimide (CF-BMI) and quartz-fiber-reinforced bismaleimide (QF-BMI) composites can meet the structure-function integration requirements of next-generation aviation equipment, and the structural design of their bonding zones directly affects their service performance. Hence, in this study, the carbon-fiber-reinforced bismaleimide composite ZT7H/5429, the woven quartz-fiber-reinforced bismaleimide composite QW280/5429, and epoxy adhesive film J-116 were used as research materials to investigate the influence of the bonding area size on the mechanical properties, and this study proposes a novel design methodology combining radial basis function (RBF) neuron machine learning with the NSGA-II algorithm to enhance the mechanical properties of the bonded components. First, a finite element simulation model considering 3D hashin criteria and cohesion was established, and its accuracy was verified with experiments. Second, the RBF neuron model was trained using the finite element tensile strength and shear strength data from various adhesive layer parameter combinations. Then, the multi-objective parameter optimization of the surrogate model was accomplished through the NSGA-II algorithm. The research results demonstrate a high consistency between the finite element simulation results and experimental outcomes for the epoxy-bonded CF/QF-BMI composite joint. The stress distribution of the adhesive layers is similar under the different structural parameters of adhesive films, though the varying structural dimensions of the adhesive layers lead to distinct failure modes. The trained RBF neuron model controls the prediction error within 2.21%, accurately reflecting the service performance under various adhesive layer parameters. The optimized epoxy-bonded CF/QF-BMI composite joint exhibits 16.1% and 11.2% increases in the tensile strength and shear strength, respectively.
碳纤维增强双马来酰亚胺（CF-BMI）与石英纤维增强双马来酰亚胺（QF-BMI）复合材料之间的环氧粘结接头能够满足下一代航空设备的结构-功能一体化要求，其粘结区域的结构设计直接影响其服役性能。因此，在本研究中，采用碳纤维增强双马来酰亚胺复合材料ZT7H/5429、编织石英纤维增强双马来酰亚胺复合材料QW280/5429和环氧胶膜J-116作为研究材料，研究粘结面积大小对力学性能的影响，并提出一种将径向基函数（RBF）神经网络机器学习与NSGA-II算法相结合的新颖设计方法，以提高粘结部件的力学性能。首先，建立了考虑三维哈希因准则和粘结力的有限元模拟模型，并通过实验验证了其准确性。其次，利用各种粘结层参数组合的有限元拉伸强度和剪切强度数据训练RBF神经网络模型。然后，通过NSGA-II算法完成代理模型的多目标参数优化。研究结果表明，环氧粘结CF/QF-BMI复合材料接头的有限元模拟结果与实验结果高度一致。尽管粘结层的结构尺寸不同会导致不同的失效模式，但在不同结构参数的胶膜下，粘结层的应力分布相似。训练后的RBF神经网络模型将预测误差控制在2.21%以内，准确反映了各种粘结层参数下的服役性能。优化后的环氧粘结CF/QF-BMI复合材料接头的拉伸强度和剪切强度分别提高了16.1%和11.2%。
