摘要
基于数据驱动的故障诊断方法已被广泛应用于旋转机械零部件故障诊断领域。目前,大多数诊断方法主要依赖于定长数据分割产生的大量数据,但分割的数据通常为短周期的小片段信号,而实际长周期冗余信号由于数据尺度不匹配,无法直接作为测试样本进行故障识别。针对以上不足,提出了一种新的基于数据概率密度与一维卷积神经网络(Data Probability Density and One-Dimensional Convolutional Neural Network,DPD-1DCNN)的故障诊断方法,其具有两个特点:①提取信号的密度特征可抵抗数据的冗余;②适应不同长度的冗余信号可作为诊断模型的输入。该方法采用DDS试验台产生的行星齿轮箱故障数据进行了验证;其在保证高诊断精度的同时,又增强了诊断模型的适应性。
Data-driven fault diagnosis methods have been widely used in the field of fault diagnosis of ro⁃tating machinery components.However,most of the current research methods mainly rely on a large amount of data generated by fixed-length data segmentation.The segmented data is usually a short-period small segment signal,and the actual long-period redundant signal cannot be directly used as a test sample for fault identifica⁃tion.In view of the above shortcomings,a new fault diagnosis method based on data probability density and one-dimensional convolutional neural network(DPD-1DCNN)is proposed.It has two characteristics:①the density feature of the extracted signal resists the redundancy of the data;②adapt redundant signals of different lengths as input to the diagnostic model.The method is verified on the planetary gearbox fault data generated by the DDS test bench,which not only ensures high diagnostic accuracy,but also enhances the adaptability of the diag⁃nostic model.
作者
张搏文
庞新宇
关重阳
Zhang Bowen;Pang Xinyu;Guan Chongyang(College of Mechanical and Vehicle Engineering,Taiyuan University of Technology,Taiyuan 030024,China;Shanxi Key Laboratory of Fully Mechanized Coal Mining Equipment,Taiyuan 030024,China)
出处
《机械传动》
北大核心
2023年第3期113-119,共7页
Journal of Mechanical Transmission
基金
国家自然科学基金(52175108)
山西省重点研发项目(202102010101009)。
