【作者】 邢宗义； 刘松； 季海燕； 秦勇； 贾利民；

【Author】 XING Zong-yi~1,LIU Song~1,JI Hai-yan~1,QIN Yong~2,JIA Li-min~2 (1.School of Mechanical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China; 2.State Key Lab of Traffic Control and Safety,Beijing Jiaotong University,Beijing 100044,China)

【机构】 南京理工大学 机械工程学院； 北京交通大学 轨道交通控制与安全国家重点实验室；

【摘要】 为解决不同乘坐舒适度标准之间无法比较和转换的问题,采用遗传神经网络技术,进行乘坐舒适度标准之间的相关性分析研究,实现了舒适度标准之间的精确建模。首先采用轨道谱和动力学仿真软件生成振动加速度信号,然后以UIC标准为例介绍了舒适度标准的计算方法,最后采用遗传神经网络构建舒适度标准之间的相关性模型。神经网络的结构采用经验法确定,其参数采用遗传算法与Levenberg-Marquardt算法的组合进行训练。仿真结果表明:乘坐舒适度标准之间具有强相关性,采用遗传神经网络可以实现舒适度标准之间的精确建模。更多还原

【Abstract】 To solve the comparison and transformation problem of different ride comfort indexes,a correlation analysis approach based on genetic neural networks was proposed.Firstly,vibration accelerations were obtained by track spectrum and ADAMS/Rail dynamic simulation software.Secondly,how to calculate ride comfort standard was illustrated using UIC513 standard.Thirdly,the correlation models of ride comfort indices were constructed using neural networks.The structures of the neural networks were determined empirically,and the parameters of the neural networks were trained by combination of genetic algorithm and Levenberg-Marquardt algorithm.The experiment results show that there are high correlations between ride comfort indices,and that the genetic neural networks can convert one index to any other index successfully and precisely.更多还原