中文摘要：

为了解决综采工作面煤矸界面探测问题,提出了利用煤矸下落冲击钢板的振动特征来探测煤矸界面的新方法。煤矸振动信号表现出非平稳特征,采用EMD方法可以将复杂矿井环境下的煤矸振动加速度信号分解成固有模态分量,每个模态分量都包含了特有的时间尺度特征。将包含煤矸振动特征的前6个IMF分量的能量,结合均值、方差及峭度等时域特征值,构成9维特征模式,作为SVM分类器的输入进行训练及分类。结果表明,基于Hilbert-Huang变换的IMF分量的能量特征能够反映煤矸振动特征的差异,SVM分类方法能够准确判断煤矸混合状态。

英文摘要：

The present paper is aimed to introduce a new method of detecting coal gangue interface in the case of a fully mechanized mining work-face by using vibrating signals of the coal gangue. Due to the non-stationary characteristic features of the response signals in the complicated environment, it is possible for us to use the empirical mode decomposition （EMD） to decompose the original vibration signals into its intrinsic mode components （IMFs）, which are characteristic of intrinsic time scale. In addition, since EMD behaves as an adaptive data-driven filter bank and can extract the signal features of the disturbance in accordance with their different physical properties. Since EMD proves to be more efficient than the conventional frequency-domain filtering waves for noise rejection, it is possible for us to use them to extract the IMFs, which can greatly help to offer us various kinds of information of the coal and gangue regardless of the noise interference and greatly improve the disturbance ratio. Moreover, the signals of interest can be clearly displayed in the time frequency domain. The IMFs contain different frequency components and denote stationary signal under the special characteristic scale. So the energy of IMFs can reflect the change of the vibration signals of the coal and gangue. And, finally, on the basis of above analysis, we have worked out the energy features of the first six IMFs and the other three time-involved parameters, such as the inputs of supporting vector machine for the simulation experiments. Furthermore, the coal-gangue interface detection is actually considered as a two-phase classification problem and the SVM is also advantageous in a two-class classification based on the search of the structural risk minimization, supported by few learning samples. Thus, our experimental results prove that the HHT technique mentioned here is highly potential in detecting vibration signals of coal and gangue and SVM can therefore be applied to classify and test the actual status of coal-gan