基于组合代理模型的汽车座椅性能多目标优化
Multi-objective optimization of vehicle seat performance based on combined surrogate model
阮志新 1陈超 2詹腾 3吴海琴 2徐从昌 4李落星4
作者信息
- 1. 湖南联诚轨道装备有限公司,湖南株洲 412001
- 2. 吉利汽车研究院(宁波)有限公司,浙江宁波 315000
- 3. 湖南联诚轨道装备有限公司,湖南株洲 412001;湖南大学机械与运载工程学院,湖南长沙 410082
- 4. 湖南大学机械与运载工程学院,湖南长沙 410082
- 折叠
摘要
基于响应面和神经网络法的组合代理模型,以模态和最大应力为优化目标,研究某款汽车主驾座椅骨架关键部件厚度的最优设计方案.首先,建立主驾座椅系统的靠背强度和模态的仿真模型,并通过标准试验对仿真模型进行精度标定;然后,选取10个座椅的关键部件厚度作为设计变量,采用最优拉丁超立方方法生产空间样本点,运用响应面和神经网络组合代理模型对样本点进行拟合预测,以1阶模态最大和最大应力最小为目标,采用NSGA-Ⅱ算法对多目标优化问题进行求解,得出关键部件厚度的最优设计方案.结果表明:在质量不增大的前提下,通过壁厚优化座椅,1阶模态提升0.19 Hz,座椅最大应力减小6.64 MPa,实现了对座椅骨架的综合性能的提升.
Abstract
In this article,based on the combined surrogate model of the response surface and the neural network method,ef-forts are made to explore a vehicle's main seat structure,in order to work out the optimal design scheme of the key components'thickness,with the modal and the maximum stress as the optimization objectives.Firstly,the simulation model of the main seat structure's backrest strength and modal is set up;the standard test is conducted to calibrate the simulation model's accuracy.Then,the thickness of the seat's 10 key components is selected as the design variables,and the optimal Latin hypercube experi-mental design method is used for space sampling.The combined surrogate model of the response surface and the neural network method is used to fit and predict the sample points.With the maximum first-order modal and the minimum maximum stress as the objectives,the NSGA-Ⅱ algorithm is used to solve the multi-objective optimization problem,and the optimal design scheme of the key components'thickness is identified.The results show that without increasing the seat's weight,the first-order modal in-creases by 0.19 Hz through wall thickness optimization,and the maximum stress reduces by 6.64 MPa,which ensures the seat structure a higher standard of comprehensive performance.
关键词
汽车座椅/多项式响应面/神经网络/多目标优化Key words
vehicle seat/polynomial response surface/neural network/multi-objective optimization引用本文复制引用
出版年
2024
NETL
NSTL
万方数据