基于改进HHT的矿山微震信号多尺度特征提取及分类研究

针对矿山微震与爆破信号难以识别问题, 提出基于改进Hilbert-Huang变换(HHT)的矿山微震信号识别方法。该方法引入互补集合经验模态分解(CEEMD)对HHT改进, 信号被自适应分解后, 计算IMF分量的偏度、峭度、Hilbert边际谱能量、Lempel-Ziv复杂度以及重构信号的分形盒维数, 运用拉普拉斯得分(LS)对5种时频域特征参数降维, 最后通过遗传算法(GA)优化的支持向量机(SVM)模型, 实现微震信号的分类识别。经400组微震和爆破信号的实例分析验证, 两类信号的5种特征参数均有较大差异, 改进HHT法识别效果优于传统经验模态分解法(EMD)和局部均值分解法(LMD), 且基于改进HHT和GA-SVM分类模型准确率达到95%, 证实了此识别方法的准确性。

Multi-scale Feature Extraction and Classification of Micro-seismic Signals Based on Improved HHT

Aiming at the difficulty in identification of micro-seismic and blasting signals in mines, a method based on the improved Hilbert-Huang transform (HHT) was proposed. With this method, the complementary ensemble empirical mode decomposition (CEEMD) was introduced to improve the HHT. After the signal was adaptively decomposed, the bias, cliff, Hilbert marginal spectral energy, Lempel-Ziv complexity and the box dimension of the reconstructed signal were calculated, and the dimension of five characteristic parameters in time-frequency domain was reduced by using Laplace-score (LS). Finally, the classification and identification of micro-seismic signals was realized with the support vector machine (SVM) model optimized by the genetic algorithm (GA). The analysis of 400 sets of actual micro-seismic and blasting signals shows that five characteristic parameters of two types of signals are quite different. The improved HHT method is superior over the traditional empirical mode decomposition (EMD) and local mean decomposition (LMD) methods in identification, and the classification models based on the improved HHT and GA-SVM can have accuracy up to 95%, which proves the accuracy of this identification method.