基于视觉测振和卷积神经网络的螺栓松动检测方法

Bolt Looseness Detection Method Based on Visual Vibration Measurement and Convolution Neural Network

螺栓松动故障的准确检测对于确保机械产品可靠性具有重要意义。为了解决现有的基于卷积神经网络(CNN)的检测方法所需的大量高质量数据难以在实际工程中获取的问题,本文提出了一种基于视觉测振和CNN的螺栓松动检测方法。通过视觉测振技术,从视频中的每一个像素点提取出振动信号,有效解决了CNN模型训练数据难以获取的问题,通过少量实测视频样本即可对CNN模型进行训练,并实现对螺栓连接状态的准确预测。本文通过一个对悬臂梁结构的敲击实验验证了所提方法的有效性。

Accurate detection of bolt loosening is significant to ensure the reliability of mechanical products.In order to solve the problem that a large amount of high quality data required by the existing detection methods based on convolution neural network(CNN)is difficult to obtain in practical engineering,a bolt loosening detection method based on visual vibration measurement and CNN is proposed in this paper.By the visual vibration measurement technology,the vibration signal is extracted from each pixel in the video,which effectively solves the problem that the training data of CNN are difficult to obtain.The CNN can be trained with a small number of measured video samples,and bolt connection conditions can be predicted accurately.In this paper,the effectiveness of the proposed method is verified by a knocking experiment on a cantilever beam structure.