基于GA-BPNN代理模型的混合式黏滞阻尼器构造优化  

Structural parametric optimization of hybrid viscous damper based on GA-BPNN surrogate model

作　　者：林拥军[1] 陈皓 郭松 LIN Yongjun;CHEN Hao;GUO Song(School of Civil Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学土木工程学院,成都610031

出　　处：《振动与冲击》2024年第9期249-260,290,共13页Journal of Vibration and Shock

基　　金：四川省自然科学基金(2022NSFSC0430);国家自然科学基金(51778532)。

摘　　要：采取CFD(computational fluid dynamics)数值模拟和基于代理模型的优化方法,研究混合式黏滞阻尼器在多参数设计空间内的构造参数优化问题。首先,分析黏滞阻尼器参数对力学性能的影响规律,基于正交试验理论选取间隙宽度、缸体内径、活塞厚度、阻尼孔直径为关键参数,利用优化拉丁超立方设计和CFD数值模拟获得初始样本点耗能性能指标;然后,使用遗传算法(genetic algorithm,GA)和反向传播神经网络(backpropagation neural network,BPNN)的方法构建代理模型,其能够准确反映构造参数耗能性能指标和构造参数之间的繁复关系。进一步地,循环更新GA-BPNN代理模型并结合非支配排序遗传算法(non-dominated sorting genetic algorithm-II,NSGA-II),以最佳耗能能力为目标,实现在全设计空间内的寻优,确定在典型阻尼介质黏度下双出杆混合式黏滞阻尼器的最优构造参数。最后,探讨了构造参数对黏滞阻尼器流动性能的影响机理。研究结果表明,优化双出杆混合式黏滞阻尼器构造参数在选用基于GA-BPNN代理模型的优化方法时,可有效提高计算效率且具有较高预测精度。从性能指标提升角度,全黏度下最优构造参数阻尼孔直径、活塞厚度、缸体内径和间隙宽度分别为1.31 mm、120.78 mm、199.87 mm、0.5 mm,相较于基准构造,最大阻尼力和能量耗材比分别提升33.8%和64.8%。该方法可为黏滞阻尼器构造参数优化及相关研究提供参考。Here,using computational fluid dynamics(CFD)numerical simulation and the optimization method based on surrogate model,structural parametric optimization problem of hybrid viscous damper was studied in a multi-parameter design space.Firstly,effects of parameters of viscous damper on its mechanical performance were analyzed.Based on the orthogonal test theory,gap width,cylinder inner diameter,piston thickness and damping hole diameter were taken as key parameters.Using the optimized Latin hypercube design(OPLHD)method and CFD numerical simulation,the energy dissipation performance index of initial sample point was obtained.Then,a surrogate model was constructed using genetic algorithm(GA)and back propagation neural network(BPNN),this model could correctly reflect complex relations among energy dissipation performance indexes of structural parameters and structural parameters.Furthermore,GA-BPNN surrogate model was circularly updated and combined with non-dominated sorting genetic algorithm(NSGA-II)to realize optimizing in the whole design space taking the optimal energy dissipation capacity as the objective,and determine the optimal structural parameters of the dual-output rod hybrid viscous damper under typical damping medium viscosity.Finally,the influence mechanism of structural parameters on flow performance of viscous damper was explored.The study results showed that optimizing structural parameters of dual-output rod hybrid viscous damper can effectively improve computational efficiency and have higher prediction accuracy when using the optimization method based on GA-BPNN surrogate model;from the perspective of improving performance indexes,under full viscosity,the optimal structural parameters of damping hole diameter,piston thickness,cylinder inner diameter and gap width are 1.31 mm,120.78 mm,199.87 mm and 0.5 mm,respectively;compared with the benchmark structure,the maximum damping force and energy to consumable material ratio increase by 33.8%and 64.8%,respectively;this method can provide a reference for

关 键 词：代理模型 黏滞阻尼器 构造参数优化 耗能性能 CFD数值模拟 

分 类 号：TU973.32[建筑科学—结构工程]