基于径向基神经网络和ASA-MMFD算法的桥式起重机主梁轻量化设计  

Lightweight Design of Main Beam of Bridge Crane Based on Radial Basis Neural Network and ASA-MMFDAlgorithm

作　　者：高林 杨瑞刚[1] 赵广立 康忠元 黄双云 徐格宁[1] GAO Lin;YANG Ruigang;ZHAO Guangli;KANG Zhongyuan;HUANG Shuangyun;XU Gening(School of Machinery and Electronics Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China;Shanxi Special Equipment and Inspection Institute,Taiyuan 030024;Chongqing Agricultural Mechanization School,Chongqing 402160,China;Zhuzhou Tianqiao Crane Co.,Ltd,Zhuzhou Hunan 412100,China)

机构地区：[1]太原科技大学机械工程学院,太原030024 [2]山西省特种设备监督检测研究院,太原030024 [3]重庆市农业机械化学校,重庆402160 [4]株洲天桥起重机股份有限公司,湖南株洲412100

出　　处：《机械设计与研究》2020年第4期171-177,共7页Machine Design And Research

基　　金：“十三五”国家重点研发计划资助项目(2017YFC0805703)。

摘　　要：为了实现桥式起重机主梁快速轻量化设计,提出了基于径向基神经网络代理模型和ASA-MMFD算法的桥式起重机主梁轻量化设计方法。在最优超拉丁立方试验设计的基础上,利用径向基神经网络代理模型建立桥式起重机截面设计参数和最大应力、最大位移和质量之间的映射关系,引入“全局+局部”的组合策略,通过自适应模拟退火算法对代理模型进行全局寻优,采用修正可行方向法进行局部搜索。利用测试函数和NASA减速器对所提方法进行验证,结果表明:在相同精度范围内,与ASA-MMFD算法相比,所提方法在调用模型次数方面大幅度减少,提高了优化效率。在此基础上,将其运用在桥式起重机主梁轻量化设计中,从而验证所提方法的适用性。In order to realize the rapid and lightweight design of the main girder of the bridge crane,a lightweight design method of the main girder of the bridge crane based on the radial basis neural network proxy model and the ASAMMFD algorithm is proposed.Based on the optimal super-Latin cubic test design,the radial basis neural network proxy model is used to establish the mapping relationship between the design parameters of the bridge crane and the maximum stress,maximum displacement and mass,and the combination strategy of"global+local"is introduced.Through the adaptive simulated annealing algorithm,the proxy model is globally optimized,and the modified feasible direction method is used for local search.The proposed method is validated by the test function and NASA reducer.The results show that compared with the ASA-MMFD algorithm,the proposed method reduces the number of calling models by 50%and greatly improves the optimization efficiency.On this basis,it is applied to the lightweight design of the main girder of the bridge crane to verify the applicability of the proposed method.

关 键 词：径向基神经网络代理模型 自适应模拟退火 修正可行方向法 

分 类 号：TB112[理学—数学]