摘要
针对如何从多维度特征中提取滚动轴承性能退化信息,构建性能退化因子的问题,提出一种基于DCNN-BiLSTM的混合输入网络。首先利用连续小波变换和24个典型时域频域特征计算公式分别得到滚动轴承振动的二维图像数据和一维时间数据,之后将两种不同维度的数据分别输入至混合输入网络进行训练,然后输入测试集数据得到滚动轴承的性能退化因子,最后利用单调性、预测性、鲁棒性对得到的性能退化因子进行评估。试验结果证明,混合输入网络结合DCNN和Bi-LSTM的优点,可有效提取滚动轴承性能退化信息,得到的性能退化因子综合效果较好。
Aiming at how to extract the performance degradation information of rolling bearings from multi-dimensional features and construct performance degradation factor,this paper proposes a hybrid input network based on DCNN-BiLSTM.Firstly,the two-dimensional image data and one-dimensional time data of rolling bearing vibration are obtained by using continuous wavelet transform and 24 typical time-domain and frequency-domain characteristic calculation formulas,respectively.Then,the data of two different dimensions are input into the hybrid input network for training,and then the performance degradation factor of rolling bearing is obtained by inputting the test set data.Finally,the monotonicity,predictability and robustness are used to evaluate the obtained performance degradation facto r.The test results show that the hybrid input network combined with the advantages of DCNN and Bi-LSTM can effectively extract the performance degradation information of rolling bearings,and the comprehensive effect of the obtained performance degradation factor is satisfying.
作者
徐峥
梁伟阁
谈芳吟
朱启瑞
Xu Zheng;Liang Weige;Tan Fangyin;Zhu Qirui(Institute of weapon engineering,Naval University of Engineering,Wuhan 430033,China)
出处
《船电技术》
2023年第1期27-32,共6页
Marine Electric & Electronic Engineering
关键词
性能退化因子
滚动轴承
深度学习
混合网络
rolling bearing
performance degradation factor
deep learning
hybrid network
