• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2024, Vol. 60 ›› Issue (24): 35-44.doi: 10.3901/JME.2024.24.035

• 仪器科学与技术 • 上一篇    下一篇

扫码分享

基于图像相位与广义曲波变换改进的旋转机械振动测量研究

张文笛1,2, 李鸿光1,2, 周吉文1,2, 王锦鸿1,2, 李振平3   

  1. 1. 上海交通大学振动、冲击、噪声研究所 上海 200240;
    2. 上海交通大学机械系统与振动全国重点实验室 上海 200240;
    3. 北方车辆研究所 北京 100072
  • 收稿日期:2023-12-18 修回日期:2024-09-14 出版日期:2024-12-20 发布日期:2025-02-01
  • 作者简介:张文笛,男,1995年出生,博士研究生。主要研究方向为视觉测振与模态辨识。E-mail:wd_zhang@sjtu.edu.cn;李鸿光(通信作者),男,1972年出生,博士,教授,博士研究生导师。主要研究方向为动力学分析与控制、信号处理和故障诊断和可靠性和寿命评估技术。E-mail:hgli@sjtu.edu.cn
  • 基金资助:
    国家重点研发(2022YFF0605600)和国家自然科学基金(12025403, U1930202, 12104155)资助项目。

Research on Improved Vibration Measurement of Rotating Machinery Based on Image Phase and General Curvelet Transform

ZHANG Wendi1,2, LI Hongguang1,2, ZHOU Jiwen1,2, WANG Jinhong1,2, LI Zhenping3   

  1. 1. Institute of Vibration, Shock and Noise, Shanghai Jiao Tong University, Shanghai 200240;
    2. State Key Laboratory of Mechanical System and Vibration, Shanghai Jiao Tong University, Shanghai 200240;
    3. China North Vehicle Research Institute, Beijing 100072
  • Received:2023-12-18 Revised:2024-09-14 Online:2024-12-20 Published:2025-02-01

摘要: 随着数码相机和数据存储技术的发展,基于视觉的测量技术引起了研究学者的广泛关注。最近,一种基于图像相位的运动估计算法凭借亚像素级别精度、高分辨率感知能力且无需表面预处理等优势得到了迅速的发展。然而,图像多尺度分解、空间频率估计等关键技术仍存在诸多不足之处。因此,在基于相位运动估计算法的基础上,提出一种改进的基于相位运动提取算法以提升计算精度。首先,使用广义曲波变换代替传统复数可控金字塔对视频中连续帧图像进行空间多尺度分析,沿径向和周向将图像分解成具有更紧凑空间支撑的频率响应子带;其次,结合O’Shea优化策略,使用双滤波算法估计和优化空间频率;再次,采用相位差和估计的空间频率计算结构的时域振动信号;最后,对转子系统进行了恒定转速和升速工况下动力学响应位移提取试验,从时域和频域上将本方法提取结果与电涡流传感器和现有方法结果进行对比,验证了所提出方法的准确性。

关键词: 视觉测振, 多尺度分解, 广义曲波变换, 双滤波, 旋转机械

Abstract: With the development of digital camera and data storage technology, vision-based measurement technology has attracted extensive attention of researchers. Recently, a phase-based motion estimation algorithm has been developed rapidly by advantages of its strong adaptability, sub-pixel accuracy and high spatial resolution. However, several key technologies, such as image multiscale decomposition and spatial frequency estimation, still have shortcomings. Thus, on the basis of phase-based motion estimation algorithm, an improved phase-based motion estimation algorithm is developed to improve the computation accuracy. First, general curvelet transform is used to perform image multiscale analysis for the images in video, and the images are decomposed into frequency domain response subbands with more compact spatial support along radial and circumferential directions; Secondly, combined with the O’Shea optimization strategy, the double filter algorithm is used to estimate and optimize the spatial frequency; Again, the time domain displacement signals are calculated using the phase difference and the estimated spatial frequency; Finally, the rotor system is tested for dynamic response displacement extraction under constant and rise speed, and the results of this method are compared with the results of eddy current sensors and existing methods in the time and frequency domains, which verifies the accuracy of the proposed method.

Key words: visual vibratory, multiscale decomposition, general curvelet transform, double filtering, rotating machinery

中图分类号: