基于太赫兹时域光谱技术的煤岩识别方法研究

煤岩识别一直是制约煤矿无人化开采的关键问题之一。传统的人工采煤因为工作环境极其复杂，很难精准地找到煤岩的分界面，容易造成欠切割或过切割现象。太赫兹光谱技术作为一种无损探测技术，能够反映出被测物体的物理和化学信息，可以成为研究煤岩识别的有效方法。采用太赫兹时域光谱技术与多元统计法—聚类分析（CA）和主成分分析（PCA）相结合的方法来识别不同种类的煤岩。通过透射式太赫兹光谱仪获得六种煤岩样品的太赫兹光谱，对其进行FFT等一系列数学计算可以得到各种样品的折射率、吸收系数以及介电常数。计算结果表明不同种类的煤岩在折射率、吸收系数上都存在差异。分析各类煤炭样品的折射率和吸收系数与样品的各组成成分含量之间的关系，可以发现碳含量是影响其样品折射率大小的因素之一，灰分含量是影响其样品吸收系数大小的因素之一。聚类分析中两类样品的欧氏距离与主成分分析中的第一主成分（PC1）得分都能反映煤岩样品之间的相似性和相异性，并且CA与PCA的结果保持一致。分别将各类样品在0.5～2.5 THz频率范围内的折射率、吸收系数与CA和PCA结合，组成太赫兹数据与煤岩之间的模型。分析表明：根据不同样品之间的相似性，两种模型中六种煤岩样品均被分为两类；在各种样品的吸收系数与CA-PCA组成的模型中，四种煤炭被聚集在一起，并且石英砂岩（GSR-4）具有很好的独特性：石英砂岩拥有最小的PC1得分值以及石英砂岩与第二类之间的欧氏距离最大，为219.03。由此可见采用太赫兹技术与多元统计方法结合，可以实现煤岩的准确识别，识别准确率可以达到100%。

Study on Coal-Rock Identification Method Based on Terahertz Time-Domain Spectroscopy

Coal-rock identification is one of the key problems restricting unmanned coal mining. Because of the extremely complicated working environment, the traditional manual coal mining is difficult to find the interface of coal and rock accurately, which is easy to cause the phenomenon of undercutting or overcutting. As a non-destructive detection method, Terahertz spectroscopy can reflect the physical and chemical information of the object under test and be an effective method to study the identification of coal and rock. In this paper, the terahertz time-domain spectroscopy and multivariate statistical method-cluster analysis (CA) and principal component analysis (PCA) are used to identify different types of coal and rock. The THz spectra of six coal and rock samples are obtained by transmission terahertz spectrometer. FFT and other mathematical calculations can obtain various samples’ refractive index, absorption coefficient and dielectric constant. The results show differences in the refractive index and absorption coefficient of different types of coal and rock. By analyzing the relationship between the refractive index and absorption coefficient of various coal samples and the content of each component of the samples, it can be found that carbon content is one of the factors affecting the refractive index of the samples, and ash content is one of the factors affecting the absorption coefficient of the samples.The Euclidean distance of two kinds of samples in cluster analysis and the score of PC1 in principal component analysis can reflect the similarity and dissimilarity between coal and rock samples, and the results of CA and PCA are consistent. The refractive index and absorption coefficient of various samples in the 0.5~2.5 THz frequency range are combined with CA and PCA to form a model between terahertz data and coal and rock. According to the analysis,the six types of coal samples in the two models can be divided into two types based on the similarity between different samples. In the CA-PCA model with the absorption coefficient of various samples adopted, four kinds of coal are clustered together. Moreover, quartz sandstone (GSR-4) has a unique characteristic: quartz sandstone has the smallest PC1 score value, and the Euclidian distance between quartz sandstone and the second type is the largest, up to 219.03. It can be seen that the combination of terahertz technology and multivariate statistical method can realize the accurate identification of coal and rock, and the recognition accuracy can reach 100%.