China Mechanical Engineering ›› 2022, Vol. 33 ›› Issue (11): 1324-1335.DOI: 10.3969/j.issn.1004-132X.2022.11.009

Previous Articles     Next Articles

Intelligent Fault Diagnosis Method for Rotating Equipment Based on Full Vector Enhanced Deep Forest

JIANG Wanlu1,2;LI Man1,2;ZHANG Peiyao1,2;ZHAO Yapeng1,2;ZHANG Shuqing3   

  1. 1.Hebei Provincial Key Laboratory of Heavy Machinery Fluid Power Transmission and Control,Yanshan University,Qinhuangdao,Hebei,066004
    2.Key Laboratory of Advanced Forging & Stamping Technology and Science(Yanshan University),Ministry of Education of China,Qinhuangdao,Hebei,066004
    3.School of Electrical Engineering,Yanshan University,Qinhuangdao,Hebei,066004

  • Online:2022-06-10 Published:2022-07-01

基于全矢增强深度森林的旋转设备智能故障诊断方法

姜万录1,2;李满1,2;张培尧1,2;赵亚鹏1,2;张淑清3   

  1. 1.燕山大学河北省重型机械流体动力传输与控制重点实验室,秦皇岛,066004
    2.燕山大学先进锻压成形技术与科学教育部重点实验室,秦皇岛,066004
    3.燕山大学电气工程学院,秦皇岛,066004
  • 通讯作者: 李满(通信作者),男,1997年生,硕士研究生。研究方向为机电系统故障诊断与预测性维护。E-mail:li_m_bz@163.com。
  • 作者简介:姜万录,男,1964年生,教授、博士研究生导师。研究方向为现代检测技术与信号处理、控制理论与控制工程、故障诊断与智能信息处理、冷轧工艺及厚度控制。E-mail:wljiang@ysu.edu.cn。
  • 基金资助:
    国家自然科学基金(51875498);河北省自然科学基金(E2018203339,F2020203058

Abstract: A new fault diagnosis method for rotating equipment was proposed based on deep forest improved by full vector data fusion to solve the problems that traditional intelligent diagnostic methods required expert knowledge and complex feature extraction, deep neural network model had complex structure and was difficult to construct, and the single-channel signals were incompleteness. According to the characteristics of vibration signals of rotating equipment, the full vector data fusion technology was combined with multi-grained scanning of deep forest to receive the homologous double-channel signals, so as to enhance the completeness of data. At the same time, the cascade layer of deep forest was improved to reduce the deep feature disappearance and feature redundancy. In order to verify the effectiveness of the proposed method, two fault diagnosis experiments of rolling bearing and axial plunger pump were carried out respectively. The results show that this method achieves a good diagnostic effectiveness on different rotating equipment and end-to-end fault diagnosis may be realized. In addition, it is also very excellent in fault recognition accuracy on small training data sets.

Key words:  , fault diagnosis, rotary machine, deep forest, full vector spectrum, data fusion

摘要: 针对传统智能诊断方法需要专家知识和复杂特征提取,而深度神经网络模型复杂度高、构建难度大,以及单源信号信息不完备等问题,提出了一种新颖的全矢数据融合增强深度森林的旋转设备故障诊断方法。该方法根据旋转设备振动信号的特点,选择全矢谱技术与深度森林多粒度扫描相结合,用于接收同源双通道信号输入,增强了数据的完备性,并通过改善深度森林级联层来减少深层特征消失和特征冗余。为了验证所提出方法的有效性,分别进行了滚动轴承与轴向柱塞泵两例故障诊断实验研究,结果表明,该方法在不同旋转设备上都有很好的诊断效果,并可以实现端到端故障诊断。此外,该方法在小训练数据集上的故障识别准确率也非常高。

关键词: 故障诊断, 旋转机械, 深度森林, 全矢谱, 数据融合

CLC Number: