田淑侠,马江东,李广棵,等.基于改进粒子群算法的层状复合材料损伤定量识别[J].河南理工大学学报（自然科学版）,2026,45(6):147-156.

Tian S X, Ma J D, Li G K, et al.Damage quantitative identification of laminated composite materials based on improved particle swarm optimization algorithm[J].Journal of Henan Polytechnic University(Natural Science) ,2026,45(6):147-156.

基于改进粒子群算法的层状复合材料损伤定量识别

田淑侠1, 马江东2, 李广棵1, 杜文辽1, 秦志辉1, 王顺强1

1.郑州轻工业大学 河南省高端装备智能制造重点实验室，河南 郑州  450002；2.河南省锅炉压力容器检验技术科学研究院，河南 郑州  450016

摘要: 目的 为解决层状复合材料损伤识别时存在的精度低、效率差、依赖基准模型等瓶颈问题，提出两种基于改进粒子群算法的损伤定量识别方法。  方法 方法一采用分阶段递进策略，构建两阶段优化机制解决高维变量导致的效率问题，首先基于差分法快速定位潜在损伤单元，将优化变量维度缩减90%以上，大幅降低计算复杂度，提升计算效率；其次利用改进粒子群算法，通过自适应惯性权重调整和动态学习因子机制，增强局部搜索能力与收敛稳定性，实现损伤的精确定位与定量计算。方法二针对方法一依赖健康模型参数的局限性，构建自主式单算法检测体系，提出面向离散化结构的种群初始化策略，使初始种群更接近真实损伤解空间，加速收敛进程，并结合边界反弹策略，动态聚焦高概率损伤区域，避免全模型搜索的资源浪费，实现仅需损伤结构单源数据即可解耦损伤特征。  结果 数值仿真和实验测试结果表明，所提两种方法均能对损伤实现准确的定位识别和定量计算，且损伤程度仿真计算误差均保持在0.5%及以下。实验测试中，方法一多次检测结果较稳定，方法二因受测试噪声、仿真模型与实际结构之间的差异性等因素影响，损伤程度量化结果存在明显波动。  结论 方法对比显示，随着变量数目增大，方法二的收敛次数显著高于方法一。因此方法一更适用于健康数据完备的快速检测场景，而方法二在缺乏基准模型时更具工程适用性。

关键词:损伤识别;层状复合材料;粒子群算法;振动响应参数;定量检测

doi:10.16186/j.cnki.1673-9787.2025090026

基金项目:国家自然科学基金资助项目（52475172）；河南省科技攻关项目（252102241039，232102231004）；河南省高校科技创新团队支持计划项目（25IRTSTHN024）

收稿日期:2025/09/18

修回日期:2026/03/20

出版日期:2026/10/09

Damage quantitative identification of laminated composite materials based on improved particle swarm optimization algorithm

Tian Shuxia1, Ma Jiangdong2, Li Guangke1, Du Wenliao1, Qin Zhihui1, Wang Shunqiang1

1.Henan Provincial Key Laboratory of Intelligent Manufacturing of High-End Equipment， Zhengzhou University of Light Industry， Zhengzhou  450002， Henan， China；2.Henan Research Institute of Boiler and Pressure Vessel Inspection Technology， Zhengzhou  450016， Henan， China

Abstract: Objectives To address the bottlenecks in damage identification of laminated composite materials， such as low accuracy， poor efficiency， and dependence on reference models， two quantitative damage identification methods based on improved particle swarm optimization （IPSO） are proposed.  Methods Method Ⅰ. A progressive two-stage strategy is adopted. First， a differential-based approach is used to rapidly locate potential damaged elements， reducing the optimization variable dimension by over 90% and significantly decreasing computational complexity. Then， an improved PSO（IPSO） algorithm with adaptive inertia weight adjustment and dynamic learning factor mechanisms is adopted to enhance local search capability and convergence stability， enabling accurate damage localization and quantification. Method Ⅱ. addressing the limitation of Method Ⅰ which relying on a reference model， a self contained single algorithm detection system by introducing a population initialization strategy tailored to discretized structures it constructed， making the initial population closer to the true damage solution space and accelerating convergence. A boundary rebound strategy is also incorporated to dynamically focus on high-probability damage regions， avoiding resource waste from full-model search. This allows damage characterization to be decoupled using only data from the damaged structure.  Results Numerical simulations and experimental tests demonstrate that both methods achieve accurate damage localization and quantification， with simulation errors of damage extent maintained below 0.5%. In experimental tests， Method Ⅰ yields relatively stable results across multiple detections， while Method Ⅱ shows noticeable fluctuations in damage quantification due to factors such as measurement noise and discrepancies between the simulation model and the actual structure.  Conclusions Comparison of the two methods reveals that as the number of variables increases， the convergence iterations of Method Ⅱ are significantly higher than those of Method Ⅰ. Therefore， Method Ⅰ is more suitable for rapid detection scenarios with complete reference data， whereas Method Ⅱ offers greater practical applicability when reference models are unavailable.

Key words: damage identification; laminated composite materials; particle swarm optimization; vibration response parameters; quantitative detection