摘要
为满足特种无人车辆高可靠机动、低成本研制等要求,在特种无人车辆的桁架车身结构设计早期阶段引入结构轻量化思想。为实现车身结构轻量化,提出基于长短期记忆网络的设计优化方法(LSTM-DO),利用LSTM-DO的快速搜索、高精度优化能力解决车身结构设计优化中收敛慢、局部最优等问题。建立了某无人车辆桁架车身参数化模型与有限元分析模型,采用代理模型技术提高设计优化过程中车身结构性能评估速度,结合LSTM-DO优化方法快速准确地生成方案。对比了常用的梯度优化算法与启发式优化算法,所提LSTM-DO方法在最优方案性能、收敛速度和鲁棒性方面均展现出明显的优势。
To satisfy the requirements of the unmanned vehicle such as high reliability inmaneuverability and low development costs,the idea of lightweight structure was involved in the early design process of the truss structure design optimization.To realize the lightweight body structure design,a Long Short-Term Memory Network based Design Optimization(LSTM-DO)method was proposed for solving the slow convergence speed and local optimization problems with its fast-searching and high-precision optimization abilities.A parameterized model and a Finite Element Analysis(FEA)model of the unmanned vehicle truss structure were constructed,which used surrogate model to accelerate the evaluation speed of the car body structure,and realized the rapid and accurate truss structure program generation by combining with the LSTM-DO method.Compared to the commonly used Gradient-Based Algorithms(GBA)and Evolutionary Algorithms(EA),the proposed LSTM-DO method offered significant advantages in terms of solution performance,convergence speed and robustness.
作者
贾良跃
郝佳
商曦文
李作轩
阎艳
JIA Liangyue;HAO Jia;SHANG Xiwen;LI Zuoxuan;YAN Yan(Industrial and Systems Engineering Laboratory,Beijing Institute of Technology,Beijing 100081,China;Yangtze Delta Region Academy,Beijing Institute of Technology,Jiaxing 314019,China;Key Laboratory of Industry Knowledge&Data Fusion Technology and Application,Ministry of Industry and Information Technology,Beijing Institute of Technology,Beijing 100081,China;China North Vehicles Researching Institution,Beijing 100072,China)
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2023年第10期3317-3330,共14页
Computer Integrated Manufacturing Systems
基金
国家重点研发计划资助项目(2021YFB1714500)。
关键词
桁架车身
结构轻量化
优化设计
长短期记忆网络
truss structure
lightweight structure
design optimization
long short-term memory network
