基于GA-BP神经网络的转子轴承系统轴承等效刚度辨识  

Identification of Equivalent Stiffness of Bearings in Rotor Bearing Systems Based on GA-BP Neural Network

作　　者：冯科伟 李锋 鲁志文 周多龙 蒋阳 FENG Kewei;LI Feng;LU Zhiwen;ZHOU Duolong;JIANG Yang(Key Laboratory of Metallurgical Equipment and Control,Jointly Established by the Ministry of Education and the Provincial Government,Wuhan University of Science and Technology,Wuhan 430081,China;Wuhan Heavy Duty Machine Tool Group Corporation,Wuhan 430081,China)

机构地区：[1]武汉科技大学冶金装备及其控制省部共建教育部重点实验室,武汉430081 [2]武汉重型机床集团有限公司,武汉430081

出　　处：《组合机床与自动化加工技术》2024年第5期126-130,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(51905388)。

摘　　要：针对转子-轴承系统的力学模型中轴承等效刚度难以确定的问题,提出了基于GA-BP代理模型的转子系统轴承参数辨识的方法。首先,建立转子-轴承系统有限元模型并进行模型的静态和动态验证;其次,搭建试验台进行模态锤击试验,得到其前四阶弯曲固有频率;最后,基于转子系统仿真模型生成GA-BP神经网络代理模型完成轴承等效刚度的辨识并做误差分析。结果表明,采用该方法可有效的识别轴承的等效刚度,并且GA-BP神经网络的识别效果优于传统的BP神经网络,其最大误差为1.52%,证明了该方法的可行性。Aiming at the problem that the equivalent stiffness of bearings is difficult to determine in the mechanical model of rotor-bearing system,a method for identifying bearing parameters of rotor system based on GA-BP proxy model is proposed.Firstly,the finite element model of the rotor-bearing system is established and the static and dynamic verification of the model is carried out;Secondly,a test bench is built for modal hammering test to obtain the natural frequency of the first four bending orders;Finally,based on the rotor system simulation model,the GA-BP neural network surrogate model is generated to identify the equivalent stiffness of the bearing and do error analysis.The results show that the equivalent stiffness of the bearing can be effectively identified by this method,and the recognition effect of the GA-BP neural network is better than that of the traditional BP neural network,and its maximum error is 1.52%,which proves the feasibility of the proposed method.

关 键 词：转子-轴承系统 参数识别 模态分析 ANSYS GA-BP神经网络 

分 类 号：TH133.3[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]