当前位置: X-MOL 学术Mech. Syst. Signal Process. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Vibration measurement of a rotating cylindrical structure using subpixel-based edge detection and edge tracking
Mechanical Systems and Signal Processing ( IF 7.9 ) Pub Date : 2021-09-17 , DOI: 10.1016/j.ymssp.2021.108437
Aisha Javed 1 , Hyeongill Lee 2 , Byeongil Kim 3 , Youkyung Han 1
Affiliation  

Acceleration sensors are commonly used for measuring the vibrations of structures. However, these contact-type sensors cannot be installed in some areas, such as on objects located in hazardous areas. Recently, non-contact-type measurement approaches, including photogrammetry techniques (e.g., point tracking, digital image correlation, and target-less approaches) have been introduced using images obtained from cameras. Nevertheless, photogrammetric approaches, e.g., point tracking and digital image correlation, have the same problem because targets or high-contrast speckle patterns need to be mounted on structures. Instead, the target-less approaches for vibration measurement were developed and tested on static structures like bridges and other civil structures. However, analysts have rarely focused on the rotating axis of cylindrical structures, which is a general component of the rotation-based renewable power generation system. Therefore, in this paper, we introduced a subpixel-based vibration measurement method for a cylindrical rotating structure based on the video images acquired from a non-contact sensor. The frames were magnified, and subpixel-based edges were detected in each frame of the video. Then, using the proposed edge tracking technique, the coordinates of the edges in a region of interest were tracked throughout the video for measuring the vibrations. The proposed edge tracking technique keeps the track of the edge locations in the previous frame as well as the locations in the pixels of the current frame. To show the effectiveness of the proposed method, two simulation datasets and one real dataset were constructed. For the simulation datasets, we generated videos by adding sinusoidal noises together with random noise in an image that contains a static cylindrical object. For the real dataset, a video of a rotating cylindrical object was acquired. The results obtained using the proposed method were compared with the results obtained using the existing multi-interval second-order differential edge detection technique and partial area-based technique. From the simulation datasets, vibrations related to both the single and multiple frequencies were effectively detected by applying the proposed method. The proposed method had the lowest root mean square error (RMSE) calculated with the reference data compared to the existing methods. In the real dataset, we could demonstrate that the proposed method could effectively detect the vibrations on the rotating axis of a cylindrical structure with the exact locations of the edges while removing the non-interest edges or false edges. Moreover, during the frequency analysis, the peaks of the proposed method results were at the same frequency at which the object was rotating. Therefore, the proposed method can be a useful solution to detect the vibration of rotating structures located in hazardous areas with uneven backgrounds and uneven brightness.



中文翻译:

使用基于亚像素的边缘检测和边缘跟踪的旋转圆柱结构的振动测量

加速度传感器通常用于测量结构的振动。但是,这些接触式传感器不能安装在某些区域,例如位于危险区域的物体上。最近,使用从相机获得的图像引入了非接触式测量方法,包括摄影测量技术(例如,点跟踪、数字图像相关和无目标方法)。然而,摄影测量方法,例如点跟踪和数字图像相关,有同样的问题,因为目标或高对比度散斑图案需要安装在结构上。相反,振动测量的无目标方法是在静态结构(如桥梁和其他土木结构)上开发和测试的。然而,分析家很少关注圆柱结构的旋转轴,它是基于旋转的可再生能源发电系统的通用组件。因此,在本文中,我们基于从非接触式传感器获取的视频图像,介绍了一种基于亚像素的圆柱旋转结构振动测量方法。帧被放大,并在视频的每一帧中检测到基于亚像素的边缘。然后,使用所提出的边缘跟踪技术,在整个视频中跟踪感兴趣区域中的边缘坐标以测量振动。所提出的边缘跟踪技术保持对前一帧中边缘位置以及当前帧像素中位置的跟踪。为了证明所提出方法的有效性,构建了两个模拟数据集和一个真实数据集。对于模拟数据集,我们通过在包含静态圆柱形物体的图像中添加正弦噪声和随机噪声来生成视频。对于真实数据集,获取了旋转圆柱形物体的视频。将使用所提出的方法获得的结果与使用现有的多区间二阶微分边缘检测技术和基于局部区域的技术获得的结果进行比较。从仿真数据集中,通过应用所提出的方法,可以有效地检测到与单频和多频相关的振动。与现有方法相比,所提出的方法使用参考数据计算的均方根误差 (RMSE) 最低。在真实数据集中,我们可以证明所提出的方法可以有效地检测圆柱结构旋转轴上的振动,并具有精确的边缘位置,同时去除不感兴趣的边缘或假边缘。此外,在频率分析期间,所提出的方法结果的峰值与物体旋转的频率相同。因此,所提出的方法可以成为检测位于背景不均匀和亮度不均匀的危险区域中的旋转结构振动的有用解决方案。

更新日期:2021-09-17
down
wechat
bug