Abstract: Ultrasonic guided wave method is an effective method to achieve large-scale bolt connection state detection,but the accurate detection model is hard to construct caused by the difference of probe positions.To realize the accurate extraction and classification of connection state features,a bolt connection state detection method combining Empirical Mode Decomposition (EMD) energy entropy and Support Vector Machine(SVM) was proposed,which combined EMD energy entropy and SVM to deal with the obvious nonlinear characteristics in ultrasonic guided wave detection and realize effective information extraction and accurate status monitoring.EMD was used to extract several levels of Intrinsic Mode Function (IMF),and the energy entropy was used to characterize the difference in the probe position.To eliminate the influence of ultrasonic probe position,a probe position recognition model based on SVM was further established,then the bolt preload force was identified through support vector regression,and the bolt preload identification model based on the fusion of EMD energy entropy and SVM was realized.The average relative error of the six sets of bolt preload identification tests was 165%,which verified the effectiveness of the detection method.

Key words: empirical mode decomposition, energy entropy, support vector machine, bolt preload, ultrasonic guided wave

摘要： 超声导波方法是实现大范围螺栓连接状态检测的有效手段,而探头位置的差异导致难以建立精确的识别模型。为实现连接状态特征的准确提取与分类,提出一种融合经验模态分解能量熵与支持向量机的螺栓连接状态检测方法,该方法针对超声导波检测中明显的非线性特性,将经验模态分解能量熵和支持向量机相结合,实现了有效的信息提取和精确的状态监测。首先采用经验模态分解技术提取各级固有模态函数,通过能量熵表征探头位置差异性;为消除超声探头位置的影响,进一步建立基于支持向量机的探头位置识别模型,并在此基础上,通过支持向量回归,对螺栓预紧力进行识别,最终建立起将经验模态分解能量熵与支持向量机融合的螺栓预紧力识别模型。试验结果显示6组螺栓预紧力识别试验平均相对误差为165%,从而验证了所提识别模型的有效性。

关键词: 经验模态分解, 能量熵, 支持向量机, 螺栓预紧力, 超声导波