基于对抗生成网络的滑动轴承支撑转子动平衡  

Balancing of Sliding Bearing Supported Rotor Based on Generative Adversarial Network

作　　者：曹俊灵 钟顺[1] 龙菲 韩佳杰 王超 

机构地区：[1]天津大学力学系,天津 [2]河北瑞兆激光再制造股份有限公司,河北 唐山

出　　处：《声学与振动》2020年第2期66-76,共11页Open Journal of Acoustics and Vibration

摘　　要：转子动平衡是转子系统检修和出厂所必要的基本步骤。本文基于对抗生成网络的思想旨在解决转子测试信号数量较少特征单一的问题,将神经网络技术应用于转子动平衡过程中。结果表明,利用生成数据进行训练的神经网络可有效的预测不平衡大小和位置,进行动平衡,特别是对于非线性支撑的转子系统,神经网络动平衡的效果要优于传统线性方法(影响系数法)。Rotor balancing is a basic step for maintenance and delivery of rotor systems. Based on the adversarial generative network, we try to solve the limitation of the small quantity and monotony of the data, and apply neural network technology to the rotor balancing process. The results show that using the generated data, the neural network can effectively predict the size and position of the imbalance. Especially for the nonlinearly supported rotor system, the neural network dynamic balancing is better than the traditional linear method (influence coefficient method).

关 键 词：动平衡 转子系统 对抗生成网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]