【摘要】 针对岩石破裂过程中声发射信号难以识别的问题，提出一种改进变分模态分解(VMD)去噪与多特征融合的声发射信号识别方法。首先，根据奇异值理论确定VMD分解模态数K,实现信号的自适应分解。然后，利用排列熵筛选出最优本征模态分量(IMF)并重构，进而提取重构信号的多个特征向量并融合。最后，将融合特征向量输入到多元宇宙算法优化的支持向量机(MVO-SVM)模型实现声发射信号的识别。实验结果表明，相比同种特征提取方法，基于改进VMD去噪的识别效果优于EEMD去噪和小波阈值去噪；相比单一特征向量，声发射信号的融合特征向量可以取得更高的识别准确率。更多还原

【Abstract】 To handle the problem of acoustic emission signal recognition in the process of rock fracture, an acoustic emission signal recognition method based on improved variational modal decomposition(VMD) denoising and multi-feature fusion is proposed in this paper. Firstly, the VMD decomposition mode number K was determined according to the singular value theory to realize the adaptive decomposition of the signal. Then, the optimal intrinsic mode function(IMF) was selected by permutation entropy, and the multi-feature vector are extracted from the reconstructed signal and fused for further recognition. Finally, the fusion feature vector was input into the support vector machine model optimized by multi-verse optimizer(MVO-SVM) to realize the recognition of acoustic emission signals. The experimental results show that, comparing with the same feature extraction method, the recognition effect of the improved VMD denoising is better than those of EEMD denoising and wavelet threshold denoising. Compared with the single feature vector, the fused feature vector of acoustic emission signal can achieve higher recognition accuracy.更多还原