摘要
提出了一种模型修正方法,可以在不依赖模型灵敏度的前提下,利用较少的计算量实现对结构有限元模型的参数修正。该方法首先构建代理模型替代结构有限元模型,通过计算少量样本点,训练支持向量回归机(support vector regression, SVR)预测参数所对应的响应;其次,以结构固有频率的残差为目标函数,利用粒子群优化算法实现全局寻优求解,得到修正后的有限元模型参数;进一步,以带孔平板为试验研究对象,基于实测数据验证了所提方法的有效性,并讨论不同参数、样本点数等对模型修正精度的影响;最后,用某卫星结构模型修正算例证明了该方法相对基于灵敏度分析的方法在计算耗时上的优势。该研究旨在为具有复杂参数-响应特征的结构模型修正提供技术支持。
Here,a model modification method was proposed,it could realize parametric modification of structural finite element model using less computational amount without relying on model sensitivity.Firstly,a surrogate model was constructed to replace structural finite element model,a small number of sample points were calculated to train a support vector regression(SVR)machine,then the machine was used to predict responses corresponding to parameters.Secondly,the residual of structural natural frequency was taken as the objective function,the particle swarm optimization(PSO)algorithm was used to realize global optimization solving,and obtain modified finite element model parameters.Furthermore,a perforated plate was taken as the test study object,the actual test data were used to verify the effectiveness of the proposed method.Effects of different parameters and sample points on the model correction accuracy were discussed.Finally,a satellite structural model modification example was used to demonstrate advantages in computing time of the proposed method over the method based on sensitivity analysis.The study results provided a technical support for modification of structural models with complex parameter-response characteristics.
作者
何子豪
吴邵庆
HE Zihao;WU Shaoqing(School of Civil Engineering,Southeast University,Nanjing 211189,China;Jiangsu Engineering Research Center of Aerospace Machinery,Southeast University,Nanjing 211189,China;Jiangsu Key Laboratory of Engineering Mechanics,Southeast University,Nanjing 211189,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2023年第15期163-172,240,共11页
Journal of Vibration and Shock
基金
江苏省优秀青年基金(BK20180062)
江苏省“六大人才高峰”项目(KTHY-005)。
关键词
有限元模型修正
支持向量回归机(SVR)
SVR响应面
粒子群优化
试验研究
finite element model updating
support vector regression(SVR)machine
SVR response surface
particle swarm optimization(PSO)
test study
