基于悬挂式FBG的螺栓连接微小扭矩检测

Small Torque Detection of Bolt Connection Based on Suspended-FBG

将光纤布拉格光栅通过尾纤悬挂在带螺栓连接的结构件表面,使用结构动态检测法测试螺栓连接结构的扭矩。测试中,在待测结构件上产生周期性的振动,在振动的作用下尾纤中产生周期性应变,它作为波源在刻有布拉格光纤光栅的光纤中形成应力波,布拉格光纤光栅感知应力波形成测试信号。在识别过程中,首先使用经验模态分解方法对测试信号进行分解以去除不平稳分量及噪声,然后提取信号的有量纲和无量纲特征值,最后将这些特征值输入基于支持向量机的识别系统。结果表明,该方法识别正确率达97.2%,扭矩识别能力在N·cm量级。本研究开拓了一种新的螺栓连接状态的检测方法,尤其适用于小质量结构中微小扭矩的检测。

Loosing of the bolt connection structure affects its working and operation safety.The main reasons for the loosing coming from loading,vibration,and friction.Consequently,loosing is inevitable and its monitoring is important for its application.At present,the theory and technology of testing the tightness state of bolt connection are still not mature.The detection of small torque is a technical difficulty in this field.In this study,for the structure to be tested on the uneven surface in the narrow space,the Bragg Fiber Grating(FBG)was used as the sensor to identify the small torque of the bolt connection.In the testing,a periodic vibration in the tested structure with bolt tightness information was excited and employed for the identification.One tail of the suspended-FBG was sticked on the tested structure,and the vibration yielded periodic strains in the tail of the FBG,which acted as the source of the elastic longitudinal wave propagating along the optical fiber with a FBG written in it.The edge-filter method was used to demodulated the signals in the FBG sensor to satisfy the high frequency signals.The information coming from the FBG was used to be identified.Firstly,the Empirical Mode Decomposition(EMD)method was used to decompose the original signal,basing on which,we removed the unstable components and noise by calculating the correlation function of each component and the original signal.Then the signals were restructured for later identification.The dimensional features(standard deviation,residuals,peak-peak value,and energy)and dimensionless features(skewness,kurtosis,waveform factor,amplitude factor,impact factor and margin factor)of the signals were exacted,and were inputted to the recognition system based on the Support Vector Machine(SVM)finally,where we used the ten-fold cross-validation algorithm and Gaussian kernel function SVM for higher accuracy.Results show that the recognition accuracy reaches to 97.2%and the torque recognition ability is on the order of N·cm.This study proves that optical fiber is a good acoustic waveguide,and the installation technique of suspended FBG effectively mitigates spectral distortion resulting from uneven stress due to direct adhesion,thereby decreasing the complexity associated with sensor installation.At the same time,because the optical fiber as an acoustic waveguide does not sense the torsional displacement,the bending stress wave cannot form an effective transmission in the optical fiber,the FBG only senses the vibration displacement along the optical fiber axis that causes the longitudinal wave.Therefore,the signal deviation caused by the excitation and sensor setting in the actual test process is relatively small and limited,which can reduce the difficulty of the signal processing.On the other hand,the study identifies that the signal processing and identification method are suitable for the non-linear,non-stationary and small sample test data in this study.This study presents a new detection method for the bolted state,especially for the detection of small torques in small mass structures on the uneven surface in the narrow space.