Page 36 - 《爆炸与冲击》2026年第10期
P. 36

第 46 卷    第 10 期                  爆    炸    与    冲    击                      Vol. 46, No. 10
                2026 年 10 月                   EXPLOSION AND SHOCK WAVES                          Oct., 2026

               DOI:10.11883/bzycj-2025-0327


                          基于数据驱动混杂点阵结构的惯性效应

                                     和结构效应竞争机制分析                                  *


                                            徐    蓉,胡    钰,丁圆圆,李科斌,周风华

                                   (宁波大学冲击与安全工程教育部重点实验室,浙江 宁波 315211)

                  摘要: 混杂点阵结构通过将不同微结构胞元在空间中进行混杂排布,可实现宏观力学性能的改善,是一种有效的
               轻质点阵结构设计方法。目前该类结构的微结构设计多集中于准静态力学性能的提升,而在高速冲击载荷下的相关
               研究较为缺乏,主要原因在于惯性效应会显著影响结构的动态响应,增加多目标协同设计的复杂度。本文中以                                       2  种不
               同相对密度的菱形十二面体胞元结构构建了正交各向同性的四阶双密度混杂点阵结构,采用双向长短期记忆网络建
               立预测模型,扩展有限元数值模拟所获得的应力-应变曲线数据规模,并通过多维度评价指标验证模型预测的可靠
               性。基于准静态平面压缩分析了混杂点阵结构的比吸能行为及相应的结构效应;进一步将压缩速度提高至                                        100 m/s
               的动态加载条件,对比研究了高速冲击下惯性效应对混杂点阵结构动态响应的影响机制。根据混杂点阵结构中胞元
               混杂程度的不同,分类讨论惯性效应与结构效应之间的竞争与协同关系。通过机器学习方法实现了有限元数据的高
               效扩展,揭示了动态加载下惯性效应与胞元配比、空间排布之间的内在联系,对混杂点阵结构在抗冲击防护等工程领
               域的应用具有指导意义。
                  关键词: 混杂点阵结构;机器学习;能量吸收;惯性效应;动态力学响应
                  中图分类号: O383   国标学科代码: 13035   文献标志码: A


                   Analysis of the competing mechanisms between inertial and structural
                                 effects in data-driven hybrid lattice structures

                                   XU Rong, HU Yu, DING Yuanyuan, LI Kebin, ZHOU Fenghua

                          (Key Laboratory of Impact and Safety Engineering, Ministry of Education, Ningbo University,
                                                Ningbo 315211, Zhejiang, China)


               Abstract:  Through spatial arrangement of different microstructures, well-designed hybrid lattice structures exhibit enhanced
               mechanical properties compared with uniform lattice structures composed of a single unit cell, thereby serving as a promising
               strategy  for  lightweight  lattice  structure  design.  However,  current  microstructure  design  approaches  for  hybrid  lattices
               primarily  focus  on  improving  quasi-static  compressive  mechanical  performance.  In  reality,  engineering  protection  often
               involves complex dynamic impact environments rather than simple quasi-static loads. However, research on the microstructural
               design of hybrid lattices under high strain rate impact loading remains limited. This is due to the significant influence of inertial
               effects  on  dynamic  structural  mechanical  response,  which  complicates  multi-objective  optimization.  A  fourth-order  hybrid
               orthogonal isotropic lattice structure was constructed using two types of rhombic dodecahedron (RD) unit cells with different
               relative densities. A bidirectional long short-term memory (Bi-LSTM) network was employed to develop a predictive model
               that  expanded  the  dataset  of  stress-strain  curves  obtained  from  finite  element  simulations.  The  reliability  of  the  model’s




                 *   收稿日期: 2025-09-29;修回日期: 2026-04-22
                   基金项目: 国家自然科学基金(11702152, 12202217);浙江省自然科学基金(LMS25A020009);宁波市科技创新              2025  重大
                          专项(2022Z202);宁波大学研究生科研创新项目(IF2024026)
                   第一作者: 徐 蓉(1998- ),女,博士研究生,2401090049@nbu.edu.cn
                   通信作者: 丁圆圆(1987- ),男,博士,副教授,dingyuanyuan@nbu.edu.cn


                                                         101501-1
   31   32   33   34   35   36   37   38   39   40   41