【摘要】 为了增强相位敏感光时域反射仪(Φ-OTDR)系统对不同扰动事件的识别能力，提出了一种基于快速自适应多元经验模态分解(FA-MVEMD)样本熵和非洲秃鹫算法优化支持向量机(AOVA-SVM)的光纤振动信号模式识别算法。光纤振动信号经FA-MVEMD分解后得到若干多元本征模态函数(IMF),通过计算与原始信号的相关系数，筛选出相关系数大于0.1的有效分量并计算其样本熵作为信号的特征，最后利用AOVA-SVM分类模型进行识别。实验结果表明，相比于其他SVM分类模型，AOVA-SVM的分类准确率更高，对6种事件的平均分类准确率达到了97.5%,实现了Φ-OTDR系统模式识别的既定目标，具有实用价值。更多还原

【Abstract】 In order to enhance the recognition ability of phase sensitive optical time domain reflectometer(Φ-OTDR) system for different disturbance events, a mode recognition algorithm of optical fiber vibration signal based on fast and adaptive multivariate empirical mode decomposition(FA-MVEMD) sample entropy and African vulture algorithm optimized support vector machine(AOVA-SVM) was proposed. The fiber vibration signal was decomposed by FA-MVEMD to obtain several multivariate intrinsic mode functions(IMF). By calculating the correlation coefficient with the original signal, the effective components with correlation coefficient greater than 0.1 were selected and the sample entropy was calculated as the signal feature. Finally, the AOVA-SVM classification model was used for recognition. The experimental results showed that the classification accuracy of AOVA-SVM was higher than other SVM classification models, and the average classification accuracy of six events reached 97.5%, which realized the established goal of the pattern recognition of Φ-OTDR system and had practical value.更多还原